Introduction

Despite improvements in overall survival after diagnosis, breast cancer (BC) remains a major cause of death worldwide, with more than 627,000 estimated deaths in 2018 [1]. Recent reports have shown increases in age-specific and stage-specific BC incidence in the US [2]. However, the annual increase in BC, particularly in women < 50 years old [2], cannot be explained by increasing lifespan or changes in parity, a long-established risk factor for BC [3]. Thus, changes in other risk factors and environmental exposures may be related to these secular trends [4].

Air pollution is considered the world’s most massive single environmental health risk [5]. Air pollution contains suspended particles or particulate matter (PM) as well as secondary contaminants. In 2013, PM was classified as carcinogenic to humans and a causative agent of lung cancer [6]. Further, the ESCAPE study showed a clear association between PM10 and PM2.5 exposure and incident lung adenocarcinoma [7]. This evidence is consistent with the fact that lungs are the organ most directly affected by air pollutants. Yet, extensive evidence describes systemic effects of air pollution, particularly low-grade systemic inflammation and oxidative stress—both mechanisms linked to cancer progression [8].

Higher air pollution may be linked with increased cancer mortality. For example, long-term (1-year) PM2.5 exposure was associated with all-cancer mortality, mainly from BC in women and lung cancer [9]. Similarly, a Japanese study revealed a higher BC mortality [10] as well as in the US [11]. These findings suggest that long-term PM exposure during carcinogenesis and once cancer is established may have direct and indirect effects contributing to cell proliferation, invasiveness, and potentially metastasis, including in BC. However, the specific contributions of PM2.5 to BC worse outcomes have not been determined. This study aimed to determine the specific contributions of long-term PM2.5 on breast cancer patients.

Methods

Population

The cohort included patients with locally advanced breast cancer and advanced breast cancer treated at the National Cancer Institute in Mexico City. Patients were recruited from 2012 to 2015. Inclusion criteria were individuals who had histopathological confirmation of stage I–IV breast adenocarcinoma, were ≥ 18 years of age, lived in Mexico City for at least 1 year before diagnosis, and had not received previous cancer treatment. The exclusion criterion was histopathological confirmation of a diagnosis different from adenocarcinoma. The total eligible population included in the study was 151 women and they were followed up until March 2020 (5–8 years of follow-up). Socioeconomic status was obtained from the social worker's initial review based on income and home characteristics and classified as follows: level 1 (lowest stratum) to level 5 (highest stratum).

Determining long-term PM2.5 exposure before BC diagnosis

Long-term (1-year average) PM2.5 exposure was established using high-resolution satellite aerosol optical depth measurements for 1 × 1 km2 grids [12]. The daily prediction model used calibrations using municipal ground monitors (≥ 18 h available collected on Tapered Element Oscillating Microbalance devices), land use, and meteorological features. Mexico City Atmospheric Monitoring System (SIMAT) provided meteorological stations and networks, including temperature and humidity data for the final prediction model. The exposure for each patient was obtained using their home address from 1 year before diagnosis. Our study included the region of Mexico City between the longitudes of − 99.42 to − 98.67 W and latitudes of 19.09–19.77 N. To avoid the mountainous uninhabited area to the southeast and east of the city, we delimited the study area to the valley adjacent to Mexico City.

Clinical and histopathological characteristics of BC patients

We obtained and registered all available clinical characteristics (e.g., tumor size, stage, body mass index, lymph node status) in research datasets. Histopathological assessment of the initial biopsies was done by expert pathologists and included type, grade [according to the Nottingham classification using the Scarff–Bloom–Richardson (SBR) scale as a surrogate and obtained for every sample included], presence of inflammatory cells, presence or absence of in situ components, lymphovascular invasion, perineural invasion, and other parameters. Cell proliferation markers Ki67 (anti-Ki67, RTU, clone MIB-1; Dako North America, CA, USA) were evaluated using standard immunohistochemical protocols. In brief, in full histologic sections, at least three high-power (× 40 objective) fields were selected to represent the spectrum of staining seen on initial overview of the whole section. We use the invasive edge of the tumor and only nuclear staining was considered positive and counting at least 500 malignant invasive cells. Three independent observers assessed the immunostaining. Cell proliferation index was calculated according to the presence of Ki67 as the ratio of positively stained nuclei to total nuclei in malignant cells.

Statistical analyses

For descriptive analyses, we determined the measures of central tendency and variability of all variables included. We examined correlation between PM2.5 levels and the values of continuous variables. Subsequently, multivariable-adjusted analysis was performed using logistic (i.e., triple-negative phenotype) and linear (i.e., tumor size at diagnosis), to determine the associations between PM2.5 exposure and the characteristic of worse prognosis. We adjusted the models for potential confounders, including age and body mass index (as a continuous variables), socioeconomic status [low (levels 1 and 2) vs. high (levels 3, 4, and 5)], family history of BC (present vs. absent), smoking status (positive vs negative, anytime before diagnosis), and histologic grade [low and intermediate (SBR < 8) vs. high grade (SBR ≥ 8)]. We also tested for linearity of the associations for continuous variables and explored the role of categorized age (> 50 vs. ≤ 50 years) and menopause (present vs absent at diagnosis) in subanalyses. All analyses were run using R studio software (R Project for Statistical Computing, CRAN, The Comprehensive R Archive Network, Vienna).

Results

Population and PM2.5 levels

A total of 151 women diagnosed with BC living in Mexico City and treated at the National Cancer Institute—Mexico were included. Mean age was 51.9 years (SD 11.0 years). The population included a high frequency of overweight (BMI > 25) and obese (BMI > 30) individuals (33.8 and 38.4%, respectively), with a mean of 29.0 kg/m2 (SD 5.9 kg/m2). The frequency of comorbidity was 13.9% for type two diabetes mellitus. Mean tumor size at diagnosis was 6.4 cm, and the most frequent clinical stage was IIIA (38.7%) (Table 1). BC patients lived in heterogeneous locations within the city (Fig. 1). Long-term PM2.5 estimations indicated a mean concentration of 23.0 µg/m3 (SD 1.9 µg/m3), with a normal distribution (p = 0.11, Shapiro–Wilk Test, Supplementary Fig. 1).

Table 1 Demographic, clinical, and histopathological characteristics of patients treated at the National Cancer Institute—Mexico, from December 2012 to June 2015 (N = 151)
Full size table
Fig. 1
figure 1

Home location of patients with breast cancer who were treated. a Map and b satellite of Mexico City and. Each red dot represents where at least one patient lives

Full size image

Association between long-term PM2.5 exposure and clinical characteristics

We observed positive correlation between long-term PM2.5 exposure and tumor size at diagnosis (Pearson correlation coefficient, r = 0.22; p = 0.007) (Supplementary Fig. 2). Subsequently, a multivariable linear regression model adjusted for age, socioeconomic status, Ki67, histological grade, and smoking status indicated an independent linear association between long-term PM2.5 exposure and tumor size at diagnosis [Risk Ratio (RR): 1.39; 95% CI 1.10, 1.76; p = 0.007] (Table 2, Supplementary Figs. 2, 3). According to these results, for every increase of two standard deviations (3.88 µg/m3) in long-term PM2.5 exposure, BC tumors increased by 0.8 cm. Analysis by subgroups did not show a clear effect on younger (RR 1.31; 95% CI 0.99, 1.73; p = 0.06) but a significant effect on postmenopausal women (RR 1.49; 95% CI 1.04, 2.13; p = 0.028). No significant association was observed related to clinical stage (p = 0.94), or node status (p = 0.61).

Table 2 Unadjusted and multivariable-adjusted linear regression model for the association between long-term (1-year average) PM2.5 exposure and tumor size in patients with breast cancer, treated at the National Cancer Institute—Mexico December 2012 to June 2015 (N = 151)
Full size table

Association between long-term PM2.5 exposure and histopathological characteristics

We did not find a significant association with long-term PM2.5 exposure (p = 0.87) or Ki67 levels (p = 0.09). We did not observe any significant association regarding pathologic complete response (pCR) in the subset of patients (n = 100) who received neoadjuvant chemotherapy (p = 0.63). Five intrinsic BC subtypes are characterized based on estrogen and progesterone receptors (ER, PR) and the human epidermal growth factor receptor 2 (HER2-enriched). Compared with other subtypes (luminal A, luminal B, and HER2−enriched), triple-negative BC (TNBC; ER−/PR−/HER2−) represents around 15% of all BC and confers a worse prognosis. Long-term PM2.5 levels were not associated with luminal A (23.1%; p = 0.74), luminal B (53.6%; p = 0.28), or HER2+ phenotype (10.5%; p = 0.69). However, TNBC patients (12.5%) had a median exposure of 24.10 µg/m3 [interquartile range (IQR): 1.19 µg/m3], which was significantly higher than other phenotypes (median: 22.80 µg/m3; IQR 1.17 µg/m3; p = 0.015) (Supplementary Fig. 4). This association was evident in multivariable-adjusted models (RR 1.59; 95% CI 1.13, 2.23; p = 0.008) (Table 3) adjusted for age, family history of breast cancer, histologic grade, socioeconomic status, and smoking. Remarkably, this effect was stronger and statistically significant in younger women (RR 2.182; 95% CI 1.15, 4.14; p = 0.035). We also found a significant association in premenopausal women (RR 1.90; 95% CI 1.008, 3.58; p = 0.047). Exploring differences in molecular subtypes according to clinical stage potentially related to our finding in TNBC, we found significant differences for luminal A-like, but not for TNBC (p = 0.67, Supplementary Table 1).

Table 3 Unadjusted and multivariable-adjusted logistic regression model for the association between long-term (1-year average) PM2.5 exposure and triple-negative phenotype in patients with breast cancer, treated at the National Cancer Institute—Mexico December 2012 to June 2015 (N = 151, TBNC, n = 16)
Full size table

Cross-effect of outcomes linked to long-term PM2.5 exposure

Using sensitivity models, we explored whether significant outcomes associated with long-term PM2.5 exposure influenced the observed effects. We found that the effect of long-term PM2.5 exposure on tumor size at diagnosis was independent of molecular phenotype (i.e., TNBC), keeping consistency of the estimates and statistical association. For the association between long-term PM2.5 exposure and TNBC, we did not found changes in the estimates, and the statistical significance remained (p = 0.011) (Table 4).

Table 4 Sensitivity models for the association between long-term (1-year average) PM2.5 exposure on tumor size at diagnosis and molecular subtype in patients with breast cancer, treated at the National Cancer Institute—Mexico, from 2012 to 2015 (N = 151)
Full size table

Effect of long-term PM2.5 exposure, tumor size, and TNBC on overall survival

We did not observe any significant association between long-term PM2.5 exposure and relapse-free and overall survival (p > 0.05 for both analyses, log-rank test; Fig. 2a). We also did not find a significant association between tumor size and relapse-free and overall survival (p > 0.05). However, we observed an expected significant association between TNBC and overall survival (p < 0.001; Fig. 2b), even with a low number of deaths (n = 19) at the time of analysis.

Fig. 2
figure 2

Kaplan–Meier curves. It shows the association between a long-term PM2.5 exposure and overall, high and low exposure categories were based on the mean value (23.0 μg/m3); and b triple-negative breast cancer phenotype and overall survival

Full size image

Discussion

We observed an association between higher long-term PM2.5 exposure levels (1-year average) and larger tumors at diagnosis as well as higher risk of TNBC. Both factors are related to poor prognosis in BC patients, particularly those with TNBC. TNBC generally shows aggressive biological and clinical characteristics, including earlier onset of metastatic disease, visceral metastases, rapidly progressive disease, short response duration to available therapies, and inferior survival outcomes [13]. TNBC is a heterogeneous subtype of breast cancer that is beginning to be refined by its molecular characteristics and clinical response to a targeted therapeutic approach [14]. TNBC represents 15% of all breast cancers. However, its prevalence varies in different settings, with the highest prevalence of 39% of breast cancers in premenopausal African American women compared to 13% in all women in the US [15]. TNBC confers a poor prognosis, mainly because it cannot be treated with endocrine or anti-HER2 therapy; hence, chemotherapy appears to be the only available treatment modality [16]. Other factors linked to TNBC include higher body mass index during premenopausal years, higher parity, and lower lifetime duration of breastfeeding [17]. Probably, most relevant finding linked to the association between air pollution and TNBC phenotype is BRCA1 mutations. More than 75% of tumors arising in women carrying a BRCA1 mutation have TNBC and a basal-like phenotype (one of five intrinsic subgroups of breast cancer), although most TNBC tumors are not mutated for BRCA1 [18].

Air pollution components including polycyclic aromatic hydrocarbons as benzo(a)pyrene, 3 dioxins, sulfur-containing compounds (SO3, H2SO4), and 3-nitrobenzanthron have been shown to be mutagenic in human cells, with a clear dose–response relationship [19]. However, the direct effects of PM on epithelial cells are unlikely to induce mutagenesis. Still, indirect effects (e.g., low-grade inflammation or oxidative damage) can cause DNA damage, adducts, and mutations. Inflammation generates reactive oxygen and nitrogen species, which can help combat pathogens and stimulate tissue repair and regeneration. However, these species can also damage DNA, which, in turn, can promote mutations that initiate and promote cancer [20]. If air pollution contributes to TNBC, as suggested by our results, BRCA1 mutations might be a plausible pathway. To our knowledge, this is the first evidence of an association between air pollution and BC phenotypes, particularly TNBC.

In countries with limited economic and health resources, BC is usually diagnosed in advanced stages leading to worse outcomes including higher mortality. Further, air pollution is strongly linked to poverty [21]. Evidence suggests that nearly 92% of pollution-related deaths occur in low- and middle-income countries [22]. Our study found an association between long-term PM2.5 levels and tumor size at diagnosis, suggesting potential co-occurrence of late or absent screening in this population as well as a potential contribution to cell proliferation or inhibition of cell death facilitated by air pollution [23]. Experimental studies observed that tumor growth occurs in an exponential or Gompertzian manner, and cells die by apoptosis in each generation; some cells undergo senescence, while others are eliminated by the immune system [24]. Therefore, transformed cells need several divisions to form a macroscopic tumor, thus requiring a more extended time window than afforded by this study (1 year before diagnosis). In addition, in vitro evidence suggests that air pollution components may exert an effect on cell proliferation as well as on apoptosis and may justify our finding related to tumor size, with stronger and significant effect in postmenopausal women [25].

Hypotheses and evidence about the carcinogenic mechanism of suspended particles abound, including release of cytokines and induction of oxidative stress, activation of signaling pathways mediated by microRNA dysregulation, promotion of angiogenesis through upregulation of vascular endothelial growth factor, and p53 mutations [26]. Our findings might suggest that long-term PM exposure can generate a microenvironment favoring cancer cell proliferation, although the mechanisms underlying this potential association as well as in vitro demonstration are needed. Further studies about other factors linked to air pollution and late diagnosis (e.g., poverty, low education level) are also needed.

Mexico City is one of the largest cities in the world [27] with a population exceeding 20 million and with heavy vehicular traffic, responsible for emission of > 220 tons/month of suspended particles. The geographic characteristics of the region [28] result in PM2.5 levels higher than most US and European urban areas [12], and frequent environmental emergencies arise due to air pollution in Mexico City. This highly polluted environment poses higher risks to all citizens, but our findings suggest an effect on patients with breast cancer and potentially on other cancer patients.

Suspended particles are responsible for several harmful health effects. In urban areas, fossil fuels are the primary source of these particles, mainly from transportation, power stations, and factories. Reducing the levels of particulate contamination after sustained interventions (especially policies and regulation) is associated with improved public health, although the potential oncological effect has not been determined. In 2009, Pope et al. analyzed PM2.5 data from 51 cities. After adjusting for other risk factors, they found that reduction of PM2.5 levels during 1980–2000 was strongly associated with an overall increase of 2.7 years in life expectancy [14]. Likewise, the SAPALIDIA study conducted in Switzerland in 1991–2002 analyzed lung diseases in adults from eight communities and showed that decreased annual mean concentrations of PM10 levels by 5–6 μg/m3 attenuated the yearly rate of lung function decline. These data suggest that continued improvement of air quality is needed to decrease morbidity and mortality associated with exposure to suspended particles. Our results open exciting avenues into whether reducing air pollution may mitigate worse outcomes and mortality in cancer patients.

Our study has some limitations including limited statistical power and inability to control for some confounding by factors derived from the geographic region (Mexico City), which might be correlated with both exposure and lack of BC screening and thus with tumor size at diagnosis. Our findings may also be influenced by other factors not included in the final multivariable models (e.g., diet, educational level) because of power limitations. Other limitations of this study are that we only explored the effect of PM2.5 exposure, but ambient air pollution combines several different pollutants and mixtures, including metals, polycyclic aromatic hydrocarbons, and gases. However, this is one of the first studies conducted focusing on cancer outcomes and tumor characteristics, which lays foundations to continue exploring the effects of other environmental pollutants such as PM10, ozone, endocrine disruptors, or metals on the clinical, pathological, and molecular characteristics of BC patients. Lack of generalizability is also a concern, and our results may not be applicable to other patients. However, to study a homogeneous ethnical population so well characterized as ours, may gives light about air pollution effects that can be applicable to other ethnicities and geographical areas.

To conclude, our study provides the first evidence of an effect of long-term PM2.5 exposure (1 year before diagnosis) on specific clinical and pathological characteristics, including larger breast tumors and triple-negative phenotype. These characteristics are associated with poor prognosis in women with BC. Future prospective studies with higher statistical power in other populations are needed to confirm and generalize our findings.