Patient recruitment in the US

Community recruitment across the US was conducted from July 2011 to August 2016. For community recruitment, we designed and received IRB approval for a protocol allowing patient ascertainment through the internet, by phone or in person (NCT0151900). The study was published online in ClinicalTrials.gov and advertised on social media. We also recruited patients at community centers, such as Nueva Vida, Baltimore and Richmond (https://www.nueva-vida.org/), the Avon Breast Cancer Walk, Washington, DC (https://www.avon.com/breast-cancer-crusade), and the Oklahoma Latino Community Development Agency. In addition, we sent letters to oncologists in predominantly Hispanic areas in Southern Florida., sending letters to oncologists in personal contacts with Hispanic cancer support and business groups, and through charities.

In Texas, we recruited subjects exclusively from the Texas Tech University Health Sciences Centers at Lubbock and El Paso. We invited women receiving care for current or previous breast cancer diagnoses to participate in the study. All US and Guatemala patients were administered a questionnaire in Spanish or English. The questionnaire for participants included age of diagnosis, demographics, reproductive history, socioeconomic status, and relevant family history of breast cancer. The questionnaire was identical for Guatemalan and US subjects (Online Appendix A).

Patient recruitment in Guatemala

In Guatemala, all recruitment occurred directly within major medical centers. In Guatemala City, we recruited patients from the Hospital General San Juan de Dios (HGSJDD) and the Instituto Nacional de Cancerologia (INCAN). Both Hospitals serve the entire Guatemalan population, with patients referred by primary care physicians. Most breast cancer patients treated at INCAN were older than 40 (84%) and lived in the capital city, indicating a higher SES. Furthermore, younger patients came from regions west of Guatemala City (22%). The questionnaire for Guatemalan patients included age at the time of diagnosis, demographics, reproductive history, cooking on a wood stove (a measure of SES), and family history of breast cancer.

IRB approval and patient consent

In Guatemala this study was conducted at the Hospital General San Juan de Dios (HGSJDD) and the Instituto de Cancerología (INCAN) in Guatemala City. The Research Ethical Committees of each institution approved the protocol, and the study was determined exempt from institutional review board (IRB) approval by the NIH Office of Human Studies Research. Women attending either of these hospitals for their breast cancer diagnostic biopsies were invited to participate and gave written informed consent. Two 5 ml tubes of blood were collected and frozen at − 20 °C as well as a tumor biopsy stored in 0.5 ml of RNAlater solution at − 20 °C. Trained interviewers administered an approved questionnaire including reproductive history, family history of cancer, and socioeconomic data.

In the U.S IRB protocol was approved by the NCI IRB and is listed in clinical trials.gov (https://www.clinicaltrials.gov/ct2/show/NCT01251900) (NCT01251900).

Sequencing

We used targeted sequencing to identify pathogenic variants in blood DNA from unselected Hispanic breast cancer cases from community recruitment and from two hospitals each in Texas and Guatemala. A total of 137 blood samples from INCAN, HGSJDD, and 96 blood samples from Texas (El Paso and Lubbock) were sequenced on the NOVASeq from Illumina with the Paired-end 200 bp strategy [5]. Briefly, blood DNA (200 ng) were used to produce an adapter-ligated library the Kapa HyperPlus Kit (Roche, Indianapolis, IN) using xGen Dual Index UMI Adapters (IDT, Coralville, IA) according to Kapa-provided protocol. The resulting post-capture enriched multiplexed sequencing libraries were loaded on a NovaSeq 6000 (Illumina, San Diego, CA) and paired-end sequencing was performed using read lengths of 2 × 150 bp to an average coverage of 50×.

Variant classification

We analyzed protein truncating variants in the genes BRCA1, BRCA2, PALB2, CHEK2, and ATM, as well as pathogenic missense variants in TP53, BARD1, RAD51C, and RAD51D [6]. All genes are listed in the Supplemental Table 1.

We annotated variants using SNPNexus for targeted breast cancer genes. We performed manual validation using Integrative Genomics Viewer (IGV) (https://igv.org/app). Then placed the variants into three categories pathogenic, VUS, and benign. Pathogenic variants were further confirmed using ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/) and Varsome (https://varsome.com/).

Breast cancer subtypes in US Hispanic women

Hormone receptor data was only available in the US Hispanic subjects. We characterized the four breast cancer subtypes as follows: (1) luminal A as estrogen receptor (ER) positive, progesterone receptor (PR) positive, HER2 receptor-negative, and low Ki 67. (2) Luminal B subtype ER positive, PR positive, HER2± , and high Ki67 count. (3) HER2+ subtype is ER negative, PR negative, and HER2 positive regardless of Ki67 count. (4) Triple-negative breast cancer is ER, PR, and HER2 receptor-negative. We combined ER and PR positive, HER2 negative tumors without Ki67 data into a luminal A/B group [7].

Mammography screening

Self-reported mammography screening was available for both Guatemala and US Hispanic women. We calculated active mammography screening for patients over the age of 40. We determined the difference between the age of diagnosis and age at the first mammogram. If the difference was two years or greater, we classified the subject as receiving active screening. Patients with a first mammogram less than two years before diagnosis were classified as unscreened.

Socioeconomic status (SES)

We used self-reported income brackets in the US to estimate socioeconomic status (SES). In Guatemala, SES is difficult to determine directly due to the lack of job security. As a proxy, we used cookstove type as an indicator of SES due to the known association between wood-burning stoves and poverty in Guatemala, particularly among indigenous Mayans (https://doi.org/10.1186/ISRCTN29007942, guatemalastoveproject.org/). We confirmed that wood cookstove use is associated with Indigenous American ancestry (Supplemental Fig. 1).

Statistics

We used a two-proportion Z-test (two-tailed) to assess the difference in the percentage of patients with a family history of breast cancer, contraception use, and parity. Second, we used an unpaired t-test with Welch’s correction to ascertain the relationship between breast cancer subtypes in US patients with pathogenic and non-pathogenic variants. Third, a Chi-squared test with Yates correction examined the relationship between socioeconomic status (cookstove type) and mammogram screen usage. Finally, Fisher’s exact test assessed the relationship between Guatemalan women’s cookstove use and Indigenous American ancestry. In all calculations, a p-value of 0.05 or less was deemed significant.