We consecutively enrolled 100 women aged 18 years or older with newly diagnosed, pathologically confirmed breast cancer in a prospective longitudinal cohort study between December 2016 and October 2018. The study was approved by the Institutional Review Board at the University of North Carolina at Chapel Hill and the Malawi National Health Science Research Committee, and all participants provided written informed consent at the time of enrollment.

Women with clinically suspected breast cancer underwent core needle biopsy. There is no routine breast cancer screening in Malawi currently, and essentially all women therefore presented with clinically symptomatic disease as defined by at least one self-reported physical breast abnormality consistent with breast cancer. Tumor specimens were assessed in the KCH Pathology Laboratory by a dedicated Malawian study pathologist (T.T.). Tumor grade was determined using the modified Bloom-Richardson system [24]. All specimens were evaluated for estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2 (HER2), and Ki67 antigen by immunohistochemistry and scored according to the College of American Pathologist guidelines [25, 26]. Antibodies targeting ER (NCL-L-ER-6F11; Novocastra, Buffalo Grove, IL, USA), PR (NCL-L-PGR-312; Novocastra, Buffalo Grove, IL, USA), and HER2 (ACA342B; Biocare Medical, Pacheco, CA, USA) were used for receptor detection. ER and PR positivity were defined as ≥1% of tumor cell nuclei being immunoreactive [25]. HER2 positivity (3+) was defined as >30% of tumor cell membranes exhibiting uniform intense staining, HER2 negativity (0 and 1+ ) was defined as no staining or weak, and partial membrane staining in any proportion of tumor cells [26]. In situ hybridization (ISH) was not available to adjudicate equivocal cases (2+). Following pathologic confirmation of invasive breast cancer, demographic and risk factor data were obtained through patient interviews and medical record review. HIV status was confirmed at the time of study enrollment, along with HIV viral load and CD4 count assessment for women who were HIV-positive. Patients with HIV were initiated or maintained on antiretroviral therapy throughout the study in collaboration with HIV clinicians, according to Malawi national guidelines. Breast cancer staging was performed using physical exam, chest radiography, abdominal ultrasonography, and plain radiography as clinically indicated. Other imaging studies including computed tomography were applied selectively, due to limited public sector availability and high cost. Magnetic resonance imaging and nuclear medicine scanning were not available at KCH.

Participants received treatment according to KCH institutional standards of care, which are modeled after the National Comprehensive Cancer Network Harmonized Guidelines for SSA [27]. Women with locally advanced tumors but without evidence of distant metastases were typically offered neoadjuvant chemotherapy with a goal of achieving adequate tumor shrinkage to facilitate surgical resection. Women with operable tumors, before or after chemotherapy, underwent modified radical mastectomy. Adjuvant and neoadjuvant chemotherapy typically included four to six 21-day cycles of doxorubicin plus cyclophosphamide followed by a single-agent taxane (paclitaxel or docetaxel) for three 21-day cycles. Palliative-intent chemotherapy generally included single-agent paclitaxel or doxorubicin given in 21-day cycles. Tamoxifen was administered to ER/PR+ women. No women received HER2-targeted treatments, which are not available in the Malawi public sector.

Cohort characteristics were summarized using simple descriptive statistics. Categorical data was analyzed using Chi-squared or Fisher’s exact tests and were summarized using frequencies and percentages as appropriate. Continuous variables were summarized using median values with absolute ranges. Differences between groups were evaluated using the nonparametric Mann-Whitney test as indicated. P-values were 2-sided, and a p-value < 0.05 was considered statistically significant. Kaplan-Meier curves were used to estimate overall survival (OS) and the log-rank test was used to assess differences in survival between groups. Follow-up time was calculated from enrollment until death, loss to follow-up (LTFU), or administrative censoring on January 15, 2019. Adjusted and unadjusted Cox proportional hazards regressions were used to estimate hazard ratios (HR) and 95% confidence intervals (CIs) to determine factors associated with mortality. For our multivariable cox regression model, we selected variables based on a p-value <0.2 from the univariable analysis or high clinical significance. Our multivariable analysis included age (continuous), HIV status (reference was negative status), stage (reference was stage II), and molecular subtype (reference was luminal A). Given a high HIV infection rate in our cohort, we also explored differences in patient characteristics and outcomes stratified by HIV status. To evaluate potential bias in OS estimates resulting from LTFU, additional sensitivity analyses were performed assuming all LTFU patients died at the time of their last known contact (worst case scenario). All analyses were conducted using STATA version 15.1 (STATA, College Station, TX, USA) and GraphPad Prism version 8.0 (GraphPad, San Diego, CA, USA).

Between December 2016 and October 2018, 100 of 101 eligible women with newly diagnosed breast cancer were enrolled in the cohort (Table 1), with only one woman with pathologically confirmed breast cancer at KCH declining to be enrolled. The median age was 49 years (range 21–80 years). HIV status was determined for 99 (99%) of patients in which 19 (19%) were HIV-positive—4 (4%) were newly diagnosed with HIV at the time of enrollment and 15 (15%) had a prior HIV diagnosis and were enrolled a median of 3.1 years (range, 0.1–9.9 years) after their initial HIV diagnosis. Among the 19 HIV-positive patients, 13 had CD4 count and HIV viral load measurements performed at enrollment, with a median CD4 count of 466 cells/μL (range 101–737 cells/μL) and 10 (77%) with an HIV viral load <1000 copies/mL (Table 1). Two patients had AIDS at the time of enrollment as defined as CD4 count <200 cells/μL. Seventy-eight patients presented with a palpable mass >5cm, and 28 had clinically evident tumor ulceration. Of 91 formally staged patients, 48 (53%) presented with stage III, and 27 (30%) presented with stage IV. No patients presented with stage I. Most women had T4 tumors (n=64, 68%), and an even higher proportion had at least N1 clinical lymph node involvement (n=77, 87%).

Histologic characterization of biopsies from all 100 women revealed 35 grade II tumors and 37 grade III tumors (Table 2). Immunophenotyping revealed 49 tumors to be ER+, 40 PR+, and 24 HER2+. The Ki67 proliferation index was high in the cohort overall, evidenced by 49 women with a Ki67 staining scored of >15%. Based on immunophenotype, tumor samples were further classified into anticipated molecular subtypes based on the St. Gallen International Expert Consensus with the following results: 23 luminal A (ER/PR+/HER2−), 27 luminal B (ER/PR+/HER2+ or ER/PR+/HER2−/Ki67>15%), 17 HER2-enriched (ER/PR−/HER2+), 25 triple negative/basal-like (ER/PR−/HER2−), and 8 unclassifiable tumors (Table 2) [28].

Treatment for enrolled patients is summarized in Figure 1. Women initially underwent either upfront curative- or palliative-intent treatment based on stage at presentation and tumor characteristics. Twenty-eight women immediately began palliative-intent treatment including one patient who received palliative surgery, most of whom presented with stage IV disease. Of 72 women eligible for curative-intent treatment, four patients died before receiving treatment, three elected not to receive any cancer treatment despite counseling, and 65 received treatment. Among 65 women who received initial curative-intent treatment, 12 received upfront surgical intervention. Fifty-three women were eligible for possible delayed surgical intervention after neoadjuvant treatment with adequate tumor downsizing, 35 of whom underwent surgery (Figure 1).

As of January 15, 2019, vital status was known for 84 of the 100 women enrolled in the cohort. Median OS was 20.2 months, and 1-year OS was 74% (95% CI, 62–83%). Sensitivity analysis incorporating worst case scenario for LTFU patients resulted in median OS 16.5 months and 1-year OS was 67% (95% CI, 55–76%) (Figure 2A). Advanced stage was associated with decreased OS (Figure 2B; p<0.001), such that one-year OS for stage II was 93% (95% CI, 61–99), for stage III was 85% (95% CI, 68–93), and for stage IV was 43% (95% CI, 22–63). Larger tumor size and lymph node involvement were also associated with worse OS (Figure 2C&D; p=0.018, p=0.027). Worst case scenario sensitivity analyses revealed similar associations with mortality for stage, tumor size, and clinical lymph node involvement.

Tumor characteristics had mixed associations with mortality. There was no detected relationship between histologic grade and mortality (p=0.26), but immunophenotypically assigned molecular subtype had a significant relationship with OS (Figure 2E&F; p=0.015). One-year OS among patients with luminal A breast cancer was 100%, for luminal B was 74% (95% CI, 50%−88%), for HER2-enriched was 60% (95% CI, 22%−84%), and for triple negative was 57% (95% CI, 30%−75%).

Cox regression analyses were used to determine whether specific factors were associated with increased mortality (Table 3). Our multivariable analysis included age, HIV status, stage, and molecular subtype (Table 3). Advanced stage was associated with mortality (HR for stage IV vs. stage II, 8.86; 95% CI, 1.07–73.25; p=0.043). Compared to patients with luminal A tumors, mortality was also associated with having a HER2-enriched (HR, 7.46; 95% CI, 1.21–46.07; p=0.031), and triple-negative (HR, 7.80; 95% CI, 1.39–43.69; p=0.020) tumor immunophenotype.

In adjusted analyses, we also found HIV status was associated with increased mortality (Table 3; HR, 5.15; 95% CI, 1.58–16.76; p=0.006), although this was not observed in unadjusted analyses. Given HIV infection in 19% of our cohort, reflecting high seroprevalence overall in Malawi, we explored differences in patient characteristics and outcomes stratified by HIV status (Table 4). The median age of HIV-positive and HIV-negative women was similar, and the distribution of tumor stage and molecular subtypes were also comparable between both groups. Furthermore, HIV-positive women had a 1-year OS of 71% (95% CI, 43%−87%), and HIV-negative women had a 1-year OS of 73% (95% CI, 58%−84%) (Figure 3A; p=0.40), although exploratory analyses stratified by the presence of distance metastases at enrollment suggested possible adverse effects of HIV particularly among the subgroup of women with distant metastases (Figure 3B; p<0.001). Since immunosurveillence plays a role in breast cancer progression [29], we further evaluated whether a low CD4 count (<500 cells/μL) within our HIV-positive patients was associated with distant metastases, yet we did not observed a significant association (Fisher’s exact p=0.075).