Immunohistochemistry vs Molecular Breast Cancer Subtyping Among African Populations

Abstract

Importance Breast cancer subtyping based on immunohistochemistry (IHC) guides systemic therapy where gene expression profiling assays are unavailable, yet its validity against molecular reference standards remains poorly characterized in African populations, where laboratory infrastructure and access to pathology services are limited. Objective To evaluate concordance between IHC-based breast cancer subtyping and PAM50 intrinsic subtypes in East and West African populations and identify resource-appropriate strategies to improve classification. Design, Setting, and Participants This cross-sectional study included Kenyan (diagnosed 2012-2022) and Ghanaian women (diagnosed 2013-2015) with invasive breast cancer, from whom formalin-fixed, paraffin-embedded tumor specimens were collected during the same periods. Gene expression profiling and data analyses were performed between April 2020 and August 2026 for both cohorts. Exposures Tumor subtyping using standard IHC markers (estrogen receptor [ER], progesterone receptor [PR], and ERBB2 status and Ki-67) and histologic grade vs PAM50 intrinsic subtyping derived from a gene expression panel. Main Outcomes and Measures Concordance between IHC-based and PAM50 intrinsic subtypes was assessed using Cohen κ statistics, with sensitivity and specificity calculated against PAM50. Results Among 460 women in Kenya with breast cancer (mean [SD] age at diagnosis, 50.3 [12.5] years), 224 (49.4%) had grade 3 disease and 148 (42.3%) had stage III or IV tumors. A total of 451 women in Ghana with breast cancer (mean [SD] age, 49.9 [12.0] years; 317 [75.1%] with grade 3 disease) were enrolled as a validation cohort. PAM50 classified patients’ tumors as luminal A (201 tumors [43.7%]), luminal B (82 tumors [17.8%]), ERBB2 -enriched (50 tumors [10.9%]), basal-like (99 tumors [21.5%]), and normal-like (28 tumors [6.1%]). Concordance between IHC and PAM50 was highest for basal-like tumors (sensitivity ranging from 73.3% [95% CI, 61.9%-82.9%] to 79.6% [95% CI, 70.3%-87.1%]; specificity ranging from 96.9% [95% CI, 93.9%-98.6%] to 97.3% [95% CI, 94.9%-98.7%]) and moderate for ERBB2 -enriched tumors (sensitivity was 50.0% across all schemes, with 95% CIs ranging from 33.8%-66.2% to 35.5%-64.5%; specificity ranging from 97.6% [95% CI, 95.6%-98.9%] to 97.9% [95% CI, 95.6%-99.2%]). Luminal subtypes showed substantial misclassification using ER, PR, and ERBB2 status alone. Incorporation of Ki-67 (≥20%) or histologic grade improved discrimination between luminal A and luminal B tumors, increasing luminal B classification from 12.2% (95% CI, 8.5%-16.7%) to 41.2% (95% CI, 35.2%-47.4%), even when PR status was excluded from the panel. Findings were independently validated in women with breast cancer from Ghana. Conclusions and Relevance In this cross-sectional study of African breast cancer populations, standard IHC markers performed well for basal-like tumors but frequently misclassified luminal subtypes without proliferation markers. When testing was constrained, proliferation markers contributed more to molecular approximation than PR, with histologic grade providing useful information when Ki-67 was unavailable. These findings support tiered, resource-adapted classification where molecular assays are inaccessible.