Germline sequence variation in cancer genes in Rwandan breast and prostate cancer cases

Female breast cancer (fBC) and prostate cancer (PC) are among the most common cancers diagnosed in the world¹. In Rwanda, breast cancer is the second most common cancer among females, and prostate cancer is the most prevalent malignancy among males, according to the Rwanda National Cancer Registry, which has been hosted at the Rwanda Biomedical Center since 2018². The Rwanda registry data suggest that the age-specific incidence of breast cancer is highest between ages 55–59 years at 120 cases per 100,000 people, while prostate cancer occurs at a later age at onset and peaks at ages 70–74 at 500 cases per 100,000 (Personal communication from Dr T Maniragaba, July 2024). The global cancer observatory (GLOBOCAN), whose primary data source is the national registry, shows high mortality rates attributable to prostate cancer and breast cancer, accounting for 13% and 8.2%, respectively of the overall cancer mortality^3,4. While there are multiple contributory factors, limited public cancer awareness is a main contributor to delayed disease presentation⁵.

Several cases of fBC and PC, as well as male breast cancer (mBC) are caused by inherited pathogenic/likely pathogenic germline variants (PV). Knowledge of PV can benefit patients at risk of, or affected by cancer if appropriate genetic testing, counseling, prevention, and treatment practices can be identified^6,7,8,9,10. For example, carriers of BRCA1/BRCA2 PV are at increased risk of developing cancers at multiple sites. Germline genetic testing (GGT) is of clinical value because identifying actionable PV can lead to improved outcomes by increasing screening; reducing incidence risk with strategies such as risk-reducing surgery or chemoprevention; or tailoring effective treatment regimens^8,9. For example, poly (ADP-ribose) polymerase (PARP) inhibitors can be incorporated into the treatment pathway of select patients with germline and/or somatic PV at BRCA1/BRCA2 and other genes in the homologous recombination repair (HRR) pathway.

There is growing evidence that a significant proportion of individuals with cancer in sub-Saharan Africa (SSA) harbor PV^11,12,13,14. This work has almost exclusively been conducted in fBC cases, while limited studies in males have been undertaken to date. For example, a cohort of Nigerian fBC patients had a 2.5-fold higher rate of inherited susceptibility mutations than US Whites (2.9%), with BRCA1 being the most commonly mutated in Nigerians (7.1%)¹⁵. Similarly, an evaluation of 40 invasive fBC cases under 35 years old from the Rwanda Military Hospital and Kigali University Teaching Hospital found higher rates of BRCA1, BRCA2 and TP53 PV (2.5%, 7.5% and 2.5% respectively; overall—12.5%) compared to observations made in women with fBC of European descent¹⁶. These preliminary observations highlight the need for further large cohort, case-control and well-powered studies to better define the inheritance of BRCA1/2 and other cancer-associated PV in SSA.

There are recognized racial and ethnic disparities in GGT globally; for example, we previously reported that in the global Consortium of Investigators of Modifiers of BRCA1/2 (CIMBA) dataset as of 2018, of the over 55,000 total BRCA1/2 mutation carriers included fewer than 5% of non-White individuals¹⁷. This deficit of diverse populations in cancer genetics research limits the application of these data to all populations and leads to a large pool of variants of uncertain significance (VUS) that disproportionately impacts underrepresented populations. In addition, it is demonstrated that genetic misdiagnoses can arise when diverse populations and their mutational spectrum are not included when making inferences about pathogenicity and risk, which can exacerbate health disparities¹⁸. Thus, studies that include a broader spectrum of individuals—particularly those from historically underrepresented populations—in cancer genetics research will benefit all populations.

To address the critical gap in the diversity of reported PV and their relationship with cancer susceptibility in African populations, we undertook a case-series study of breast and prostate cases in Kigali, Rwanda.

Pathogenic Variants

Four hundred patients were approached, consented, and offered genetic counseling and testing (Fig. 1). Genetic test results were available for 342 patients (Fig. 1 and Table 1). Among fBC, 32 of 175 (18.3%) cases were observed to carry a PV (mean age at diagnosis: 38.2 years, range: 24–56 years) compared with 143 fBC cases with no PV (mean age at diagnosis: 39.8 years, range: 20–61; p = 0.057). The proportion of fBC diagnosed before age 40 was significantly higher in PV carriers than non-carriers (69% vs. 49%, two-sided Fisher’s Exact Test p-value = 0.031). 1 of 5 (20%) mBC case was observed to carry a PV (age at diagnosis: 78 years) compared with 4 mBC cases with no PV (mean at diagnosis: 47 years, range: 38–63; p = 0.157). Among PC cases, 7 of 162 (4.3%) were observed to carry a PV (mean age at diagnosis: 65.9 years, range: 59–78) compared with 155 PC cases with no PV (mean at diagnosis: 66.0 years, range: 47–84; p = 0.729). Most cases (83%) were unaware of or did not report their cancer family history.

Detailed diagram of the processes of recruitment, testing and outcomes for breast and prostate cancer patients in Rwanda.

Full size image

All fBC cases with BRCA1 (N = 11) PV had frameshift variants. Of interest, seven unrelated patients, all with the triple negative disease subtype were found with c.4065_4068del (p.Asn1355Lysfs*10), which consists in deletion of 4 base pairs in exon 10 (TCAA residues). Of the patients found with BRCA2 (N = 15), the majority had frameshift variants (N = 12), followed by large deletions (N = 2, all deletions of exon 1-4) and a nonsense mutation (N = 1). ATM mutations (N = 4) were characterized mostly with single nucleotide substitution (2 G > A; one resulting in a non-sense mutation) and two frameshift variants. TP53 PV (N = 3) were all single nucleotide substitutions (c.844 C > T, c.7306 G > A and c.1010 G > A), resulting in missense mutations. The only CHEK2 PV found was a splice donor variant (c.1461+5 G > A), while the WRN PV was a single nucleotide substitution (c.1105 C > T), and the RET PV was a frameshift variant (c.1091-1104del). Fig. 2a provides details of the PV.

a Pathogenic variants in breast cancer cases. The figure provides details about the pathogenic variants found among breast cancer cases. The majority of the pathogenic variants were frameshift deletion, followed by nonsense mutations. b Pathogenic variants in PC cases. The figure provides details about the pathogenic variants found among prostate cancer cases. The majority of the pathogenic variants were frameshift deletion, followed by nonsense mutations and intronic mutations in equal distribution. c Variants of uncertain significance in fBC cases. The figure provides details about the variants of uncertain significance found among female breast cancer cases. The majority of the pathogenic variants were missense mutations, followed by intronic mutations. d Variants of uncertain significance in mBC. The figure provides details about the variants of uncertain significance among male breast cancer cases. All variants were missense mutations. e Variants of uncertain significance in PC. The figure provides details about the variants of uncertain significance found among prostate cancer patients. The majority of the variants of uncertain significance were missense mutations.

Full size image

Among patients with PC found with BRCA2 (N = 2) PV, one had a gain (exons 17–18) and the other had a frameshift variant (c.9097dup). Two other patients were found with single RECQL4 PV, which were both frameshift variants (c.3416del and c.1048-1049del). A patient was found with a splice site variant on the BRIP1 gene (c.93+4_93+7del), while a frameshift variant was found in another patient on the RAD50 gene (c.2165 dup). Finally, a nonsense mutation was found on the RAD51D gene of a patient with PC (c.898 C > T). Table 2 provides details of the patients’ clinical characteristics and their specific PV and Fig. 2a, b provide a review of the types of variants encountered among patients with PV.

Among fBC, 30 of 32 (94%) PV were observed in genes thought to be associated with breast cancer susceptibility as classified by NCCN (Supplementary Table 2). We observed PV in genes that have been widely accepted as BC or PC susceptibility genes, including ATM, BRCA1, BRCA2, CHEK2, and TP53 (Table 2). PV not previously associated with fBC included RET and WRN ([15]). In PC, 2 of 7 (29%) PV were observed in genes thought to be associated with PC susceptibility as classified by NCCN. Both of these occurred in BRCA2. PV in BRIP1, RAD50, RAD51D, and RECQL4 were also observed, but these genes have not been previously designated as causative of PC by NCCN¹⁶.

The proportion of cancers with any PV was significantly higher in fBC (18.3%) compared with PC (4.3%; p = 0.0001; Table 1). When considering only PV that occur in genes previously known to be associated with fBC, mBC, or PC, this difference became even larger (17.1% in fBC vs. 1.2% in PC; p < 0.0001).

Of the 176 tumors observed in 175 women, 61% of fBC tumors were hormone positive (ER+ and/or PR + ), 32% were HER2 + , 26% were triple negative breast cancer (TNBC), and 16% were triple positive (Table 1). Overall fBC with PV at any locus were more likely to have been diagnosed with the TNBC disease subtype than those without a PV (49% vs. 21%, two-sided Fisher’s exact test p-value = 0.002). In addition, 10 of 11 (91%) of BRCA1 PV carriers were diagnosed with TNBC while 6 of 15 (40%) of BRCA2 PV carriers were diagnosed with TNBC (two-sided Fisher’s exact test p-value = 0.0143).

Variants of uncertain significance

Among the 342 cancer cases, N = 524 Variants of Uncertain Significance (VUS) were observed (Supplementary Table 3). Among those with no PV (N = 225, 65.8%), 461 VUS were observed. 77 cases (22.5%) had neither a PV nor a VUS. At least one VUS was observed in 257 (75.1%) cases, including 79.4% of all fBC, 60% of all mBC, and 71% of all PC cases. Figure 2c–e provide details on the found VUS and their characteristics. The most common VUS were missense variants (N = 480, 91.6%) followed by intronic variants (N = 25, 47.7%). 77 cases (22.5%) had no PV or VUS. There was no difference in age at diagnosis overall or within each diagnostic group depending on the presence or absence of a VUS. The genes that most commonly contained VUS were ATM (N = 30 VUS), BARD1 (N = 28), WRN (N = 26), and RET (N = 23), RECQL4 (N = 22), SDHD (N = 20), APC (N = 17), BRCA2 (N = 15), and PMS2 (N = 18). Similar distributions were seen in fBC (ATM, N = 14; BARD1, N = 14; WRN, N = 13; RET, N = 11; APC, N = 13; BRCA2, N = 7; SDHD, N = 9). In mBC, 6 VUS were identified among the 5 cases with one VUS each at BARD1, BLM, BRCA2, CTNNA1, DIS3L2, and RECQL4. In PC, the VUS were most commonly observed at ATM (N = 16), BARD1 (N = 13), WRN (N = 13), RECQL4 (N = 12), RET (N = 12), and SDHD (N = 11).

Our study sought to adapt well-established international eligibility guidelines to perform a pilot of clinical germline genetics testing among patients with breast and prostate cancers in Rwanda. We adapted NCCN guidelines and developed context-specific genetic counseling and testing protocols to facilitate the ascertainment of PV and VUS in this population and contribute to the characterization of inherited cancers in individuals of African descent. First, while our breast cancer study cohort was similar in age and disease characteristics to previous studies of germline genetic testing conducted from SSA countries, our study represents a large cohort of patients from a previously undertested SSA population region who underwent multi-gene panel testing^{12,13,19,20,21,22}. In our study, young BC patients with triple negative disease were the most likely to harbor PV. Second, we demonstrate that applying standard guidelines in patient selection for genetic testing with a curated multi-gene panel testing resulted in a high yield of PV across BC patients which helps to strengthen the case for more routine implementation of germline genetic testing in cancer care in SSA. Finally, our study shows a clear pattern of inheritability of the common highly penetrant genes between regions of Africa and the diaspora. Whereas BRCA2 was more prevalent in our BC cohort and many other in Eastern Africa, BRCA1 has been reported to be more common in Western Africa, Caribbean and African American women with BC^{12,15,23,24,25}.

The majority of PV identified in BC study patients (94%) were consistent with established BC risk, including ATM, BRCA1, BRCA2, CHEK2, and TP53. A high prevalence of BRCA1 or BRCA2 PV have been previously reported in unselected breast cancer cases from Nigeria, Cameroon, and Uganda^12,15,19 with BRCA1 prevalence of 6-7% and BRCA2 prevalence of 4-5%. We observed a prevalence of 6.5% for BRCA1 PV, 8.9% for BRCA2 PV, and an aggregate BRCA1/2 rate of 15.4% in the present study. Thus, our estimate of BRCA1 PV is similar to that reported in other African populations, but our prevalence of BRCA2 is somewhat higher than previously reported. In contrast, PV prevalence in unselected breast cancer case series in largely European descent populations has been estimated to be approximately 6% for BRCA1/2 in women diagnosed before age 36²⁰, 11.2% in women with triple negative breast cancer²¹, 4.7% BRCA1/2 PV prevalence in patients ages 35–64 years²². African Americans have been reported to have lower prevalence of BRCA1/2 PV compared to European Americans in the range of 4% compared to 5% in European Americans and a 1.3% BRCA1 PV prevalence in African Americans compared to 2.2% among Non-Hispanic whites²⁶. Larger African studies are needed to further establish the PV prevalence.

Two PV were observed in genes that had not been previously associated with breast cancer. One breast cancer case carried a PV in RET, a proto-oncogene that confers risks of Hirschsprung Disease 1 and Multiple Endocrine Neoplasia, Type 2 A. FGFR1, an established low penetrance breast cancer susceptibility locus, is a paralog of RET. RET exhibits elevated expression and survival in some breast tumors^27,28,29. A single PV in WRN was also identified in one breast cancer case. WRN is a RecQ-like helicase associated with (autosomal recessive) Werner and Bloom syndromes that are involved in DNA damage repair. However, no association of WRN and breast cancer have been previously reported although Werner Syndrome confers risk of sarcomas, melanoma, and thyroid cancers³⁰. Thus, while germline variation in RET and WRN are not breast cancer susceptibility genes, they each could be involved in breast cancer etiology. However, based on the pedigrees, these incidental findings do not provide evidence that the PV identified at these genes is causative. While these PV are uncommon, it is likely that their occurrence is unrelated to the cause of these malignancies.

We also observed three PV in a single breast cancer case diagnosed at age 38 with triple negative breast cancer: TP53 c.1010 G > A (p.Arg337His), BRCA2 c.3720_3723del (p.Phe1241-Valfs*17), and ATM c.7913 G > A (Table 2). ATM c.7913 G > A has been previously reported in Cameroon⁸. TP53 (p.Arg337His) has been observed in Sudan and is a founder mutation in Brazil^31,32. This PV has been reported to occur at a relatively high frequency in Li-Fraumeni Syndrome families and confers early onset breast cancer risk in Brazil^{33,34,35,36,37,38}. This PV is a founder mutation based on its occurrence on a common haplotype³⁹. The observation of this PV in Eastern Africa raises the possibility of an alternate origin of this “founder” mutation reported commonly in Brazil.

Of note, multiple identical PV were identified in fBC cases that were not known to be biological relatives. These included three occurrences of c.7913 G > A (p.Trp2638*) at ATM, seven occurrences of c.4065_4068del (p.Asn1355Lysfs*10) at BRCA1, three occurrences of c.1053del (p.Lys351Asnfs*16) at BRCA1, three occurrences each of c.3720_3723del (p.Phe1241Valfs*17) and c.5633dup (p.Asn1878Lysfs*4) at BRCA2, and two occurrences of a large deletion involving exons 1-4 at BRCA2. BRCA1 c.4065_4068del has been reported in numerous populations worldwide^40,41. This variant was identified in an earlier Rwandan cohort¹⁶ and has been uniformly reported in most East African countries and Northern Africa in patients with similar disease phenotype^11,12,23. The same variant was also reported in a large cancer genetics clinic in Norway that studied seven distinct families whose ancestries were unclear²⁵. Studies are underway to determine whether this PV reflects a distinct haplotype in Rwanda from those previously described. Similarly, c.5633dup has been previously reported in a study of Cameroonian and Ugandan breast cancer cases, and c.1053del has been reported in Uganda¹². BRCA2 c.3720_3723del has been previously reported in a set of early onset Rwandan breast cancer cases¹⁶. These observations suggest either hot spots for mutation in these genes or common (founder) PV that are relatively common in Africa. Other variants observed in this study have also been reported elsewhere. For example, BRCA1 c.1504_1508 is commonly found in Eastern and Northern Africa, but also identified in a population of unlikely African origin^11,14.

One man with breast cancer was found to be a BRCA2 carrier. While the sample size of men with breast cancer was low, the rate of PV in our cohort (20%) is within range of the prevalences seen in other populations^42,43.

BRCA2 was the most commonly reported PV among men of African descent diagnosed with PC with a frequency of 1.75%³⁰. However, we observed that Rwandan PC cases have a lower prevalence of PV than reported in other populations^{30,44,45,46,47,48}. We did not observe any PV in HOXB13, TP53, or the Lynch Syndrome genes, which have been proposed as PC susceptibility genes in other populations^48,49.

In contrast to breast cancer, the majority of PV observed in PC cases (71%) occurred in genes that are not currently thought to be PC susceptibility genes, including BRIP1, RAD50, RAD51D, and RECQL4 (Supplementary Table 2). BRIP1, RAD50, and RAD51D interact with BRCA1 and/or BRCA2 in DNA damage repair and related functions. Thus, while not known to be PC susceptibility loci, these genes are plausibly associated with PC and may warrant further evaluation in African and other populations. PV at RECQL4^50,51,52 were observed in two PC cases (c.1048_1049del and c.3416del). RECQL4 is associated with (autosomal recessive) Rothmund-Thomson syndrome, RAPADILINO syndrome, and Baller-Gerold Syndrome⁵². While this gene is located at chromosome 8q24, it is found approximately 15 Mb away from the PC susceptibility locus thought to regulate MYC in conferring PC susceptibility⁵³, and thus it is unlikely that these PV would explain long-range regulation of MYC or the association of prostate tumors at this locus.

The disparities in genomic testing led to many variants found in the population of African descent in SSA to be classified as variants of uncertain significance (VUS)⁵⁴. 115 patients (33.6%) with breast cancer and 109 patients (31.8%) with PC were observed to carry VUS. Similar prevalences have been reported in other SSA populations, congruent with findings in generally under-represented populations, highlighting the need for further larger population studies and analyses^12,19. In addition to giving a thorough explanation of non-actionable findings, we presented the VUS rates in this selected population with a suggestive phenotype in order to lay the groundwork for future studies on variant reclassification.

Our study had some limitations. First, as a cancer case series, there is limited possibility of generalizing the findings across patient populations in Rwanda and beyond. We were also unable to establish a cause-effect relationship between the variants and patient’s cancers due to the lack of a population-based non-cancer control group. Second, family history was taken during the time of encounter (pre-genetic testing counseling session) with the study participants, which was limited by recall bias and relatively poor access of the overall population to cancer diagnostic services. Third, given a regional history of migration, we could not ascertain the origins of the unrelated probands found with recurrent variants. Finally, there were practical challenges with many samples requiring retesting and a high number of participation refusals for repeat testing; this may point to potentially inadequate education during the pre-genetic testing counseling and underappreciation of burdens related to patient participation in the study. Subsequent studies will focus on cascade testing of at-risk relatives.

Our study shows a high yield of PV among high-risk individuals, when ascertained using standard guidelines. Although not validated in African populations, the NCCN guidelines provide a flexible framework to identify the population requiring cancer genetic testing based on demographic and clinical factors, even in the absence of reported family history. Further confirmatory studies are needed for validation purposes, including to determine if NCCN guidelines need to be adapted to the African setting and incidence of pathogenic variants in patients who do not meet NCCN guidelines. This work is planned in subsequent larger studies.

The World Health Organization recognizes human genomics as a means to prevent, diagnose and predict, manage, monitor, and treat genetic conditions⁵⁵. A balance needs to be maintained between approaching cancers through familial programs and evidence-based universal testing for eligible individuals and cancers, which could both prove to be innovative ways to address and curb cancer mortality in Africa. Their impact has a wide reach, in that it improves healthcare systems through human and infrastructural capacity building, while emphasizing on a precise nature of cancer prevention and treatment^54,56,57. The purpose of genetic testing is to provide individuals and systems with potentially actionable information regarding cancer. Our approach of identifying patients based on standard eligibility criteria was mostly dependent on age and disease characteristics, and less with family history, which was incomplete in most cases. In the example of Rwanda, and as illustrated on our pedigrees, familial structures and knowledge of pertinent medical history are limited. This is a common theme in most SSA-based cancer genetic studies, usually compounded with lack of follow-up, resulting in missed opportunities for familial identification, cascade testing and the implementation of intensified screening and/or risk-reducing interventions. Approaching families can help identify individuals at high-risk of cancers, and this could complement cancer screening efforts that are ongoing on the continent. Population-based screening for cancer-susceptibility genes has an advantage of identifying all carriers regardless of family history and is a proactive option that solves the failure of prevention resulting from eligible patient genetic testing. However, the adoption of this screening in low-resource settings requires a careful consideration of several factors, including the population-specific cancer risk conferred by these genes, the cancer screening infrastructural availability after genetic testing, and would be warranted in communities where a greater acceptance of cancer genetic testing exists, and a prevalence of mutational prevalence has been established⁵⁸. Our study used multi-gene panel testing, which is now the standard procedure for cancer genetics services, owing to its cost-effectiveness, speed, and efficiency, when compared to whole genome sequencing. For low resource settings however, a precise curation of genes for different disease conditions is needed, to limit the risk of incidental findings beyond the cancer scope, that these developing settings might not be able to manage.

The critical next steps will consist in understanding the study and result disclosure experience of the probands through qualitative surveys, providing a structure for disclosure of the genetic results from the identified probands to at-risk relatives, and subsequent cascade genetic testing^58,59. Effective risk communication and models of disclosure that would perform well in SSA have not been explored at length, and low health literacy could impact the probands’ intent to communicate, the information content and the motivation to undergo genetic testing^58,59,60,61. However, a shift from traditional familial disclosure and letter sharing to individualized, culturally sensitive methods has resulted in positive outcomes with regard to genetic risk assessment in women of African descent and their family members and could be tested in our population. SSA health systems differ, and methods of familial approach (physician or proband-led) could vary from a setting to another. Understanding personal values and the health-seeking behaviors of the at-risk relatives are important in establishing a relationship with families, and to facilitate the risk communication and genetic testing. Finally, health insurance schemes in Rwanda (both public and private) cover access to screening, prevention and essential cancer treatment⁵, but some mitigating strategies, including access to PARP inhibitors are not widely available (for the public insurance scheme). The current study provides a basis for future interventions, including clinical trials to test efficacy of these medications in an African population.

By using adapted eligibility guidelines, our study findings contribute to bridge the gap in the diversity of reported PV and their relationship with breast and prostate cancer susceptibility in the Rwandan population. As screening, prevention, and targeted treatment solutions emerge for individuals who carry PV, there is need for familial cancer registries for an early ascertainment, better cancer risk assessment, tailored clinical-grade laboratory solutions and diversified treatment modalities in low-resource settings.

Participant Eligibility

The first phase of this study was to identify and adapt existing clinical protocols in ascertaining eligible patients. The National Comprehensive Cancer Network (NCCN v.1. 2021)^62,63 guidelines were adapted to establish participant eligibility for both BC and PC germline genetic testing. The investigators opted to use these guidelines with minor modifications made based on the setting, the target population, and the likely ancestry of the cohort.

Eligibility criteria are outlined in Supplementary Table 1. Briefly, based on our adapted protocols, patients with Breast Cancer (BC—both fBC and mBC) were eligible for participation in this study if they were (1) diagnosed on or before the age of 45 years; (2) diagnosed under age 51 years with unknown or limited family history; (3) diagnosed under age 51 years with a second breast cancer diagnosed at any age; (4) diagnosed under age 51 years with 1 or more close blood relatives with breast, ovarian, pancreatic, or PC at any age, (5) diagnosed under age 61 years with triple-negative breast cancer (NB: At the time of the 2020 guidelines, the age cutoff for triple-negative breast cancer was 61 years, and patients with this histology subtype were recruited based on this age limit); (6) diagnosed at any age with 1 or more close blood relatives with breast cancer diagnosed under age 51 years or ovarian, pancreatic, or high-risk PC at any age; (7) diagnosed at any age with 3 or more total diagnoses of breast cancer in patient and/or close relatives.

With the same protocols, PC cases were eligible for this study if they were diagnosed with (1) metastatic, intra-ductal/cribriform histology, or high-or very-high-risk group; (2) any risk group with 1 or more close relatives with breast cancer under age 51 years or ovarian, pancreatic, or high-risk PC; (3) any risk group with 2 or more close relatives with breast or PC at any age.

Provider Education

Given the absence of trained, dedicated cancer genetic counseling staff locally at the starting time of the study, a virtual short course was developed (www.africanoncogenetics.org/home). This course was attended by over 1000 persons in the first three months of its development, including the Rwandan clinicians involved in participant accrual for this study. Rwandan clinicians also participated in in-person educational sessions in Kigali, Rwanda before study accrual began. Participating clinicians included oncologists, breast surgeons, and urologists from five teaching hospitals in Rwanda with the capacity to diagnose and/or treat cancer patients: the Rwanda Military Hospital, Kigali University Teaching Hospital, Butare University Teaching Hospital, King Faisal Hospital Rwanda, and the Butaro Cancer Center of Excellence. The referring centers were informed about the study and the eligibility criteria and contacted the study Principal Investigator to discuss each patient’s eligibility prior to enrollment.

Participant Accrual and Study Cohort

The second phase of the study was to identify and accrue study participants. Between April 2022 and April 2023, the study team assessed 400 histologically confirmed breast cancer (196 women and 5 men) and PC (199 men) patients referred from the five teaching hospitals mentioned above. All patients were seen at one institution (Rwanda Military Hospital) post referral. Figure 1 provides a schematic summary of participants accrual, eligibility, testing, and the final study sample of 342 participants with complete study data. The breakdown of patients meeting each eligibility criteria for cases included here are presented in Supplementary Table 1.

Local treating clinicians applied the inclusion criteria to potentially eligible patients and evaluated the results in collaboration with international experts in genetic cancer risk assessment.

GGT was covered by the study funding, and participants were offered transport compensation. These clinicians discussed with all the participants the study information. After this discussion, informed consent was sought and obtained from all participants prior to their involvement in the study and pre-genetic testing counseling. The counseling discussion included a communication of genetic risk based on their eligibility criteria, the genetic testing process, sample and data handling, possible results, their significance, and their potential clinical implications. A survey was then conducted to obtain participants’ demographics, confirmation of study eligibility, cancer risk history, family history along with pedigree drawing, and details on diagnosis, staging, and treatment. All consented participants provided a saliva sample collected by two trained study nurses using a collection tube obtained from Invitae Corporation (San Francisco, CA, USA). Samples were aggregated and organized in order of study participants’ enrollment and shipped to Invitae for analyses.

The Rwanda National Ethics Committee (Reference numbers: 996/RNEC/2021 and 456/RNEC/2022) and the Institutional Review Board of the Dana Farber Cancer Institute approved the study. Our study complied with the ethical regulations governing the approving Institutional Review Boards (Belmont Report and Declaration of Helsinki).

Genetic Testing

All patients underwent clinical-grade GTC using a panel of 84 cancer-predisposition genes (Supplementary Table 2). Full-gene sequencing, deletion/duplication analysis, and variant interpretation were performed at Invitae® as previously described^64,65,66. Full gene-sequencing coverage included coding exons and 10 to 20 base pairs (bp) of adjacent intronic sequence on either side of the coding exons.

Genetic variants were clinically interpreted using Sherloc⁶⁵, a refinement of interpretation guidelines from the American College of Medical Genetics and Genomics and the Association for Molecular Pathology⁶⁶ and classified as pathogenic (P), likely pathogenic (LP), variant(s) of uncertain significance (VUS), likely benign, or benign. P/LP variants and VUS were reported to the patients’ treating clinicians.

Results disclosure

The team that provided pre-genetic testing offered results disclosure to all study participants and discussed risk to family members and next steps in treatment with patients found with PV. A handout of the result was shared with the referring physician directly, or through the patient. All treating physicians had access to the study PI (AVCM) and were made aware of the ability to contact the PI for information related to clinical management of the patient. Clinicians who were involved in this study received training in genetic testing, counseling, and management developed by the research team.

Statistical Methods

Contingency tables and two-sided nonparametric statistical tests including Kruskal-Wallis (to compare differences in ages at diagnosis) and Fisher’s exact tests (FET) were generated using gtsummary to describe the distribution of PV by cancer site, clinical characteristics, and other traits. Null hypotheses were rejected when p < 0.05. Data processing was carried out in R version 4.2.2, utilizing various libraries. Initially, variants were parsed through custom R scripts and transformed into MAF formats. These formats facilitated the creation of oncoplots for both PV and VUS using the maftools package.