Automated mRNA‐Based Detection of ER, PR, HER2, and Ki‐67 in Breast Cancer Tissue and Cytology Samples in Resource‐Limited Settings

ABSTRACT

Background

Accurate detection of estrogen and progesterone receptors (ER/PR), HER2, and Ki‐67 is essential for molecular classification of breast cancer and in deciding upon tailored treatment. In many low‐ and middle‐income countries (LMICs), access to histopathology and immunohistochemistry (IHC) is limited. STRAT4, an automated RT‐qPCR assay performed on the widely available GeneXpert platform, offers a potential alternative. We evaluated the performance of STRAT4 on core biopsy, unstained smear of fine‐needle aspiration biopsy (FNAB‐US), and cellblock samples compared with conventional IHC in 178 histopathology‐confirmed breast cancer patients.

Methods

Matched core biopsy, cellblock, and FNAB‐US samples were analyzed using STRAT4. Concordance with IHC for ER, PR, HER2, and Ki‐67 was assessed using percentage agreement and kappa statistics. The effect of alternative dC _t cut‐offs used in an earlier study was also examined. Molecular subtype classification derived from STRAT4 was compared with IHC‐based classification.

Results

Concordance between STRAT4 and IHC was high for ER across all sample types (κ: 0.78–0.86). PR detection showed substantial agreement for core biopsies (κ = 0.78) and cellblocks (κ = 0.63), but only fair for FNAB‐US samples (κ = 0.53). HER2 detection showed moderate agreement (κ = 0.41–0.58), improving with modified cut‐offs (κ = 0.57–0.66), especially for FNAB‐US. Ki‐67 concordance was low (agreement 58.9%–71.0%). Kappa values indicating agreement between molecular subtype classifications using STRAT4 on various samples and routine IHC were as follows: core biopsies = 0.60; cellblocks = 0.61; FNAB‐US samples = 0.45.

Conclusion

Detection of ER from FNAB‐US, cellblocks, and core biopsies using STRAT4 showed high concordance with IHC. However, the concordance was low to moderate for other markers.

1. Introduction

Breast cancer mortality is projected to rise by 68% globally by 2050, with the highest burden expected in low‐ and middle‐income countries (LMICs) due to late diagnosis and limited access to stage‐appropriate treatment [1]. In sub‐Saharan Africa, the pooled 5‐year survival rate is only 40% [2]. To improve outcomes, the World Health Organization's Global Breast Cancer Initiative (GBCI) prioritizes timely diagnosis and access to comprehensive management [3].

Accurate molecular subtyping of breast cancer is essential to guide therapy. It relies on assessment of estrogen and progesterone receptors (ER, PR), human epidermal growth factor receptor‐2 (HER2) status, and Ki‐67 proliferation index [4]. The standard immunohistochemistry (IHC) method and fluorescence in situ hybridization (FISH) for equivocal HER2 cases require sophisticated laboratory infrastructure often unavailable in LMICs. Classification into luminal A/B, HER2‐enriched, and triple‐negative subtypes based on molecular profiles determines prognosis and treatment. Lack of reliable molecular testing hampers implementation of tailored therapy in LMICs. Alternative approaches, such as automated mRNA‐based detection of ER, PR, HER2, and Ki‐67, may offer a simplified and scalable solution for molecular profiling in resource‐constrained environments.

Xpert STRAT4 test (Cepheid, Sunnyvale, CA, USA) is a real‐time Polymerase Chain Reaction assay for semi‐quantitative measurement of ESR1 (ER), PGR (PR), ERBB2 (HER2), and MKi67 (Ki67) mRNA expression. The test is objective and technologically less demanding, requires minimal training of the technicians and is performed on GeneXpert multi‐test platform widely available in the LMICs [5]. The manufacturers recommend using formalin‐fixed paraffin‐embedded (FFPE) tissue from core biopsy specimens, which are often difficult to organize in the LMICs [6].

Using specimens obtained through fine‐needle aspiration biopsy (FNAB) offers an attractive alternative in resource‐limited settings, as its sensitivity and specificity for detecting palpable breast tumors are comparable to core‐needle biopsy [7]. However, FNAB specimens are generally considered unsuitable for conventional molecular profiling. We therefore evaluated the performance of the STRAT4 mRNA assay for detecting ER, PR, HER2, and Ki‐67 using FFPE core biopsies, paraffin‐embedded FNA cellblocks, and unstained smears of FNAB (FNAB‐US), with the objectives of assessing concordance with standard IHC.

Additionally, we assessed the concordance between STRAT4 assay performed on paraffin‐embedded cellblocks and the same assay performed on core biopsies.

2. Methods

2.1. Study Participants

The study was conducted in India at Cachar Cancer Hospital & Research Center (CCHRC), Silchar, and Post‐Graduate Institute of Medical Education & Research (PGIMER), Chandigarh. Although CCHRC is a cancer treatment center based in a semi‐urban hard‐to‐reach area of the country, PGIMER is a comprehensive tertiary care and teaching institute. At each site, women presenting with breast symptoms and having clinically and radiologically suspected breast cancers were counseled and recruited to the study after obtaining written informed consent. Patients detected to have benign conditions on subsequent core biopsy were excluded. Exclusion criteria also included age below 18 years, low performance score, previous treatment for breast cancer and having contraindications to core biopsy (e.g., on anti‐coagulant therapy).

2.2. Core Biopsy, Histopathology, and Immunohistochemistry

A core biopsy was performed under local anesthesia usually under ultrasound guidance. The formalin‐fixed tissue was processed routinely, and sections were stained with Hematoxylin and Eosin (H&E) stain for histopathology. The slides were examined by the site pathologists. Once malignancy was confirmed, the formalin‐fixed paraffin‐embedded (FFPE) tissue blocks were processed for IHC to detect ER, PR, HER2, and Ki‐67. IHC was performed on an automated immuno‐stainer (Ventana Roche system, USA in PGIMER and Epredia Autostainer‐480 in CCHRC). Rabbit monoclonal antibodies against ER (SP1 clone), PR (SP42 clone), Her2/Neu (EP3 clone), Ki67 (SP6 clone) were used and were obtained from Cell Marque, Sigma Aldrich, USA at both sites. Ultraview kit (Ventana Roche, USA) was used as the secondary antibody, followed by the chromogen. The IHC slides were initially evaluated by the site pathologists. The American Society of Clinical Oncology/College of American Pathologists (ASCO/CAP) guidelines for reporting were followed. The hormone receptors (ER/PR) were considered positive if > 1% of nuclei were stained. HER2 was scored as 3+ when > 10% tumor cells showed strong staining. A score of 0 (negative) or 1+ (weak‐positive) was considered as negative. Slides with an equivocal score of 2+ had a confirmatory FISH test. Ki‐67 index was considered high when ≥ 15% of tumor cells expressed the biomarker in their nuclei.

2.3. FNAB, Cellblock Preparation and IHC on Cellblocks

FNAB from a palpable mass was usually obtained before the core biopsy. In case of large masses or when the mass was not well delineated due to factors like edema, ultrasound‐guidance was used to place the needle. The aspirated material was expelled onto pre‐numbered glass slides and 4–8 smears were made, of which 2–4 air‐dried smears were stained by Giemsa or May‐Grünwald‐Giemsa (MGG) stain. Remaining slides were fixed immediately in 95% ethyl alcohol and stained with hematoxylin–eosin stain for cytopathological evaluation by the site pathologists. Material from each pass or from a dedicated pass was taken directly in 10% neutral‐buffered formalin for cellblock preparation. In case fluid/hemorrhagic material was aspirated, the specimen was centrifuged at 1500–2000 rpm for 10 min, and the supernatant was discarded. The cell pellet was fixed in 10% neutral buffered formalin. After fixation overnight at room temperature, the cellblock was prepared by the sodium‐alginate technique [8]. Briefly, the sedimented aspirated material was given a wash in phosphate‐buffered saline and embedded in sodium‐alginate, followed by addition of calcium chloride to form a cohesive clot. This clot was processed as a formalin‐fixed tissue specimen and embedded in paraffin to create a cellblock. Paraffin sections were cut at 3–4 μm, stained with H&E, and examined by the site pathologists for confirmation of carcinoma and adequacy of cellblock for IHC. IHC was performed on the cellblock sections for analysis of ER, PR, HER2, and Ki67 following the same procedure as described above for core biopsies and interpreted by site pathologists blinded to the IHC diagnoses on core biopsies.

2.4. STRAT4 Assay on Core Biopsy Tissues, Cellblocks and FNAB‐US

Cepheid Xpert platform was used for STRAT4 assay to detect the molecular markers from FFPE core biopsies, FFPE cellblocks, and FNAB‐US materials once breast cancer was confirmed on core biopsy. All three tests were not performed on all patients (details provided later).

Paraffin‐embedded core biopsy blocks or cellblocks were cut to obtain 3–4 sections of 4–5 μm thickness, which were placed into a 1.5 mL tube along with 260 μL of FFPE lysis buffer and 5 μL of proteinase‐K. After vortexing at maximum speed for 5 s, the mixture was incubated in a heat block at 80°C for 30 min. 260 μL of ethanol was added to the mixture and vortexed again for 15 s. For each sample, 520 μL of the prepared lysate was added directly to a test cartridge. The single‐use cartridges contained the RT‐PCR reagents for the simultaneous detection of ESR1, PGR, ERBB2, MKi67, a cytoplasmic FMR1 interacting protein 1 (CYFIP1) reference gene, an internal RT‐PCR control (CIC), and an internal Probe Check Control (PCC). The reference gene verified specimen adequacy and was used to normalize the mRNA expression levels. The assay utilized six separate fluorescent channels for target or control/reference detection, each having its own cut‐off parameters for validity.

Unstained FNAB smears (FNAB‐US) were used for the STRAT4 assay. Cellularity in the smear was assured by comparing the immediately previous smear stained by MGG stain from the same pass of FNAB. In some cases, two smears were used. Briefly, the material on an air‐dried FNAB‐US was scraped off a glass slide using a surgical blade into a 1.5 mL tube. Subsequent process was different between two sites. At PGIMER, the tube was vortexed after adding 260 μL of FFPE lysis buffer and 20 μL of proteinase‐K. After incubating at 80°C for 30 min, the sample was transferred to a 5 mL tube and 260 μL of absolute ethanol was added, and the tube was vortexed again. For each sample, 520 μL of the prepared lysate was added directly to a STRAT4 cartridge.

At CCHRC a simpler protocol was followed: 520 μL of lysis buffer was added to the sample prior to vortexing. After that the sample was micro‐centrifuged for 15 s and 520 μL of the supernatant was transferred to the cartridge. No ethanol or proteinase‐K was used.

The detection cut‐offs of CYFIP1 and CIC controls define ranges of minimum and maximum cycle threshold (C _t) values that determine a valid result, an adequate minimum sample input, and no PCR inhibition. Positive and negative results of a particular target in a channel are defined by delta C _t (dC _t) values (reference gene C _t minus target gene C _t). dC _t cut‐off values recommended by the manufacturers are as follows: ESR1 −1; PGR −3.5; ERBB2 −1; MKi67 −4. A CYFIP1 C _t value exceeding 35 indicates an invalid result. For PGR and MKi67, an indeterminate result is obtained if the dC _t values for these two markers are below the specified cut‐offs and CYFIP1 C _t is > 31.

2.5. Quality Assurance

All core biopsy, FNAB, and IHC slides were reviewed by a professor in pathology (RS) from PGIMER. If any discrepancy was observed between the review diagnosis and the site pathologist's diagnosis, the final diagnosis was arrived at through mutual consultation and consensus. The site technicians were trained to perform the STRAT4 assay by experts from Cepheid. Researchers from the International Agency for Research on Cancer were responsible for study oversight, data monitoring, and site training.

2.6. Statistical Considerations

We aimed to recruit 100 breast cancer patients at each site empirically. The IHC results from the core biopsy specimen confirmed after expert review were considered as the gold standard when evaluating the agreement and test accuracy (sensitivity and specificity) of STRAT4 on tissue, cellblocks and FNAB‐US samples. Both test accuracy and overall concordance between the IHC results and the test results from STRAT4 on tissue, cellblocks and FNAB‐US samples were presented as a percentage and its 95% confidence interval (CI). The McNemar's chi‐squared test was used to test odds ratio (in this case equal to the ratio of false positives to false negatives, that is the ratio of discordant matched pairs) is one, while the kappa statistic and its 95% CI were used to assess the agreement between any two test results comparisons. The following kappa values were used to evaluate the level of agreement: 0.01–0.20 as none to slight, 0.21–0.40 as fair, 0.41–0.60 as moderate, 0.61–0.80 as substantial, and 0.81–1.00 as almost perfect agreement. Additionally, the STRAT4 results obtained from the analysis of cellblock specimens were compared to the STRAT4 results obtained from the analysis of the core biopsy specimen. All comparisons were done based on only those with valid test results expressed as positive or negative.

We reanalyzed our STRAT4 results using a set of modified dC _t values for the markers that were used by Ng et al. in their study to evaluate STRAT4 [9]. The alternative dC _t cut‐off values are as follows: ESR1 −0.5; PGR −1.7; ERBB2 −0.25; MKi67 −2.6. These cutoffs derived from pooling of data from six independent studies using FFPE specimens were observed to improve specificity across the markers with smaller impacts on sensitivity [10]. A p‐value of 0.05 was used to infer statistical differences. All statistical analyses were performed using Stata version 19 (StataCorp LLC, College Station, TX, USA).

The study was approved by the IARC Ethics Committee (IEC 21‐12) and Cachar Cancer Hospital & Research Centre (CCHRC/IRB/012/2021). Written informed consents were obtained from all participants.

3. Results

Total 178 patients with histopathology‐confirmed breast cancer (85 at CCHRC and 93 at PGIMER) were recruited to the study. None of them had carcinoma in situ and all were epithelial cancers. Overall, 45.2% of the patients were aged below 50 years and 45.2% were at clinical Stage I/II at recruitment (data not shown).

All participants had IHC to detect ER/PR/HER2/Ki67 from core biopsy (Table 1). However, IHC results for HER2 were excluded from analysis for one patient as confirmatory FISH was not done. Ki67 result was missing for another patient.

TABLE 1.

Number recruited, number of patients tested with STRAT4 and having valid results on core biopsy, cell blocks and FNAB samples (unstained smears) and number of patients having indeterminate STRAT4 results for PR and HER2.

	Study site		Total, n (%)
	CCHRC, n (%)	PGIMER, n (%)
Number of patients recruited	85	93	178
Immunohistochemistry from core biopsy	85	93	178
STRAT4 from core biopsy
No. of patients tested	50	93	143
Invalid tests a	0 (0.0)	3 (3.2)	3 (2.1)
Valid tests	50 (100.0)	90 (96.8)	140 (97.9)
STRAT4 from cell blocks
No. of patients tested	51	60	111
Invalid tests a	1 (2.0)	1 (1.7)	2 (1.8)
Valid tests	50 (98.0)	59 (98.3)	109 (98.2)
STRAT from FNAB (unstained)
No. of patients tested	35	40	75
Invalid tests a	0 (0.0)	1 (2.5)	1 (1.3)
Valid tests	35 (100.0)	39 (97.5)	74 (98.7)
Indeterminate b STRAT4 test results for PR and Ki67 among valid samples
Valid samples for STRAT4 from core biopsy	50	90	140
PR
Results obtained	50 (100.0)	90 (100.0)	140 (100.0)
Indeterminate	0	0	0
KI‐67
Results obtained	50 (100.0)	88 (97.8)	138 (98.6)
Indeterminate	0 (0.0)	2 (2.2)	2 (1.4)
Valid sample for STRAT4 from cell blocks	50	59	109
PR
Results obtained	46 (92.0)	56 (94.9)	102 (93.6)
Indeterminate	4 (8.0)	3 (5.1)	7 (6.4)
KI‐67
Results obtained	46 (92.0)	54 (91.5)	100 (91.7)
Indeterminate	4 (8.0)	5 (8.5)	9 (8.3)
Valid sample for STRAT4 from FNAB (unstained)	35	39	74
PR
Results obtained	35 (100.0)	39 (100.0)	74 (100.0)
Indeterminate	0	0	0
KI‐67
Results obtained	35 (100.0)	39 (100.0)	74 (100.0)
Indeterminate	0	0	0

Abbreviations: ER, estrogen receptors; FNAB, fine needle aspiration biopsy; HER2, human epidermal growth factor receptor 2; Ki‐67, Antigen Kiel 67; PR, progesterone receptors.

^a Invalid if no test result could be obtained for all four target molecular types.

^b Indeterminate if dC _t for PGR or MKi67 was lower than pre‐specified cut‐offs & the CYIFIP C _t was > 31.

Table 1 shows the number of patients providing different types of samples for STRAT4 and the number having valid outcomes. STRAT4 analyses on core biopsies, cellblocks, and FNAB‐US samples were performed on 143, 111, and 75 patients, respectively. Invalid STRAT4 results were obtained from 2.1% (3/140) core biopsies, 1.8% (2/111) cellblocks, and 1.3% (1/75) FNAB‐US samples. Among the valid tests, the proportion of samples giving indeterminate results on STRAT4 for PR was none for core biopsies, 6.4% for cellblocks, and none for FNABs. Indeterminate results for Ki67 were obtained for 1.4%, 8.3%, and none for core biopsies, cellblocks, and FNAB‐US samples, respectively.

The agreement between the STRAT4 results from core biopsy, cellblock and FNAB‐US specimens and IHC from core biopsies as the gold standard for comparison is shown in Table 2. Additionally, the table describes the comparison of STRAT4 results between the two types of specimens—cellblocks and core biopsies. Based on the values shown in Table 2, we estimated the concordance, kappa values, sensitivity, and specificity of STRAT4 on various types of samples (Table 3).

TABLE 2.

Distribution of test result combinations—STRAT4 on core biopsies, cell blocks and FNAB (unstained smears) samples compared to immunohistochemistry from core biopsies; STRAT4 outcomes compared between cell block and core biopsy samples (both sites together).

Test 1	Test 2	Patients assessed (n)	Test 1+	Test 1+	Test 1−	Test 1−
			Test 2+	Test 2−	Test 2+	Test 2−
			(n)	(n)	(n)	(n)
STRAT4 on core biopsy vs. IHC
STRAT4 ER on core biopsy	IHC ER on core biopsy	140	60	1	9	70
STRAT4 PR on core biopsy	IHC PR on core biopsy	140	52	11	4	73
STRAT4 HER2 on core biopsy	IHC HER2 a on core biopsy	139	55	14	15	55
STRAT4 Ki‐67 on core biopsy	IHC Ki‐67 on core biopsy	138	80	30	10	18
STRAT4 on cell blocks vs. IHC
STRAT4 ER on cell blocks	IHC ER on core biopsy	109	47	5	6	51
STRAT4 PR on cell blocks	IHC PR on core biopsy	102	32	9	9	52
STRAT4 HER2 on cell blocks	IHC HER2 a on core biopsy	108	48	19	8	33
STRAT4 Ki‐67 on cell blocks	IHC Ki‐67 on core biopsy	100	57	12	15	16
STRAT4 on FNAB (unstained) vs. IHC
STRAT4 ER on FNAB	IHC ER on core biopsy	74	31	3	5	35
STRAT4 PR on FNAB	IHC PR on core biopsy	74	14	1	14	45
STRAT4 HER2 on FNAB	IHC HER2 a on core biopsy	73	22	21	2	28
STRAT4 Ki‐67 on FNAB	IHC Ki‐67 b on core biopsy	73	8	9	21	35
STRAT4 on cell blocks vs. STRAT4 on core biopsies
STRAT4 ER on cell blocks	STRAT4 ER on core biopsy	108	43	8	4	53
STRAT4 PR on cell blocks	STRAT4 PR on core biopsy	101	32	9	13	47
STRAT4 HER2 on cell blocks	STRAT4 HER2 a on core biopsy	108	50	16	8	34
STRAT4 Ki‐67 on cell blocks	STRAT4 Ki‐67 on core biopsy	98	62	6	19	11

Abbreviations: ER, estrogen receptors; FNAB, fine needle aspiration biopsy; HER2, human epidermal growth factor receptor 2; IHC, immunohistochemistry; Ki‐67, Antigen Kiel 67; PR, progesterone receptors.

^a One patient did not have FISH to confirm HER2 status and was excluded.

^b IHC report was not available for Ki67 for one patient.

TABLE 3.

Concordance of STRAT4 test outcomes performed on core biopsies, cell blocks, and FNAB (unstained smears) samples compared with immunohistochemistry from core biopsies and comparison of outcomes of STRAT4 test on cell block and core biopsy samples.

	Patients pairs assessed	Concordance (95% CI)	McNemar p	Kappa a (95% CI)	Sensitivity (95% CI)	Specificity (95% CI)
STRAT4 test positivity defined using the original cutoff values b and compared with immunohistochemistry performed on core biopsies
STRAT4 on core biopsies
ER	140	92.9 (88.6–97.1)	0.021	0.86 (0.69–1.02)	98.4 (91.2–100.0)	88.6 (79.5–94.7)
PR	140	89.3 (84.2–94.4)	0.118	0.78 (0.62–0.95)	82.5 (70.9–90.9)	94.8 (87.2–98.6)
HER2	139	79.1 (72.4–85.9)	1.000	0.58 (0.42–0.75)	79.7 (68.3–88.4)	78.6 (67.1–87.5)
Ki‐67	138	71.0 (63.4–78.6)	0.002	0.29 (0.14–0.45)	72.7 (63.4–80.8)	64.3 (44.1–81.4)
STRAT4 on cell blocks
ER	109	89.9 (84.3–95.6)	1.000	0.80 (0.61–0.99)	90.4 (79.0–96.8)	89.5 (78.5–96.0)
PR	102	82.4 (75.0–89.8)	1.000	0.63 (0.44–0.83)	78.0 (62.4–89.4)	85.2 (73.8–93.0)
HER2	108	75.0 (66.8–83.2)	0.052	0.50 (0.31–0.68)	71.6 (59.3–82.0)	80.5 (65.1–91.2)
Ki‐67	100	73.0 (64.3–81.7)	0.701	0.35 (0.16–0.55)	82.6 (71.6–90.7)	51.6 (33.1–69.8)
STRAT4 on FNAB
ER	74	89.2 (82.1–96.3)	0.727	0.78 (0.56–1.01)	91.2 (76.3–98.1)	87.5 (73.2–95.8)
PR	74	79.7 (70.6–88.9)	0.001	0.53 (0.32–0.73)	93.3 (68.1–99.8)	76.3 (63.4–86.4)
HER2	73	68.5 (57.8–79.1)	0.000	0.41 (0.21–0.61)	51.2 (35.5–66.7)	93.3 (77.9–99.2)
Ki‐67	73	58.9 (47.6–70.2)	0.043	0.08 (0.14–0.29)	47.1 (23.0–72.2)	62.5 (48.5–75.1)
STRAT4 test positivity defined using the alternative cutoff values c and compared with immunohistochemistry performed on core biopsies
STRAT4 on core biopsies
ER	136	89.7 (84.6–94.8)	0.002	0.79 (0.63–0.96)	98.2 (90.3–100.0)	84.0 (74.1–91.2)
PR	136	89.0 (83.7–94.2)	0.607	0.77 (0.60–0.94)	88.2 (76.1–95.6)	89.4 (80.8–95.0)
HER2	135	83.0 (76.6–89.3)	0.000	0.66 (0.50–0.82)	94.2 (84.1–98.8)	75.9 (65.3–84.6)
Ki‐67	133	70.7 (62.9–78.4)	0.522	0.33 (0.16–0.50)	76.3 (66.4–84.5)	57.5 (40.9–73.0)
STRAT4 on cell blocks
ER	103	92.2 (87.1–97.4)	0.070	0.84 (0.65–1.04)	97.7 (87.7–99.9)	88.3 (77.4–95.2)
PR	96	89.6 (83.5–95.7)	0.754	0.78 (0.58–0.98)	88.6 (73.3–96.8)	90.2 (79.8–96.3)
HER2	102	80.4 (72.7–88.1)	0.263	0.61 (0.42–0.80)	85.7 (72.8–94.1)	75.5 (61.7–86.2)
Ki‐67	94	64.9 (55.2–74.5)	0.035	0.23 (0.04–0.42)	82.1 (69.6–91.1)	39.5 (24.0–56.6)
STRAT4 on FNAB
ER	74	89.2 (82.1–96.3)	0.727	0.78 (0.56–1.01)	91.2 (76.3–98.1)	87.5 (73.2–95.8)
PR	69	72.5 (61.9–83.0)	0.000	0.35 (0.16–0.54)	90.0 (55.5–99.7)	69.5 (56.1–80.8)
HER2	73	80.8 (71.8–89.9)	0.057	0.60 (0.38–0.82)	65.6 (46.8–81.4)	92.7 (80.1–98.5)
Ki‐67	66	62.1 (50.4–73.8)	0.015	0.13 (0.08–0.35)	53.8 (25.1–80.8)	64.2 (49.8–76.9)
Compared with STRAT4 performed on core biopsies
STRAT4 on cell blocks
ER	108	88.9 (83.0–94.8)	0.388	0.78 (0.59–0.96)
PR	101	78.2 (70.2–86.3)	0.523	0.56 (0.36–0.75)
HER2	108	77.8 (69.9–85.6)	0.152	0.55 (0.36–0.73)
Ki‐67	98	74.5 (65.9–83.1)	0.015	0.32 (0.13–0.50)

Abbreviations: ER, estrogen receptors; FNAB, fine needle aspiration biopsy; HER2, human epidermal growth factor receptor 2; IHC, immunohistochemistry; Ki‐67, Antigen Kiel 67; PR, progesterone receptors.

^a Kappa values ≤ 0 as indicating no agreement and 0.01–0.20 as none to slight, 0.21–0.40 as fair, 0.41–0.60 as moderate, 0.61–0.80 as substantial, and 0.81–1.00 as almost perfect agreement.

^b Original cutoff values used for the STRAT4 test positivity are −1 for ER, −3.5 for PR, −1 for HER2 and −4 for Ki‐67.

^c Alternative cutoff values used for the STRAT4 test positivity are −0.5 for ER, −1.7 for PR, −0.25 for HER2 and −2.6 for Ki‐67.

STRAT4 showed robust concordance with IHC to detect ER for all types of specimens (92.9% on core biopsies, 89.9% on cellblocks, and 89.2% on FNAB‐US) with kappa (ranging from 0.78 to 0.86) showing substantial to near‐perfect agreement. No significant difference in sensitivity and specificity was observed for the detection of ER between different types of samples.

Concordance between IHC and STRAT4 for detection of PR was substantial for core biopsies (89.3%; κ = 0.78) and cellblocks (82.3%; κ = 0.63), but only fair for FNA‐US samples (79.7%; κ = 0.53).

For HER2, concordance ranged from 79.1% for core biopsies to 68.5% for FNA‐US samples, with kappa values of 0.41–0.58, indicating moderate agreement. STRAT4 on FNAB‐US demonstrated the lowest sensitivity but the highest specificity for HER2 detection. Concordance for Ki67 detection was generally low, ranging from 58.9% (FNA_US) to 71.0% (core biopsies) with kappa showing poor to no agreement.

Reanalysis using the dC _t cut‐offs reported by Ng et al. improved STRAT4 performance for detecting PR on cellblocks and HER2 across all sample types (Table 3). Using these revised cut‐offs, agreement between STRAT4 and IHC for HER2 detection was 83.0% for core biopsies (κ = 0.66), 80.4% for cellblocks (κ = 0.61), and 80.8% for FNAB‐US samples (κ = 0.57).

We analyzed the results of FNAB‐US samples separately for CCHRC and PGIMER as the former used a simpler protocol (Table S1). No significant difference was observed. Rather the concordance between IHC and FNA‐US to detect PR was higher at CCHRC. No major difference was observed between the sites for other types of specimens (Table S2).

STRAT4 results on core biopsies showed highest agreement to detect ER (88.9%) and lowest agreement to detect Ki67 (74.5%) when compared to STRAT4 performed on cellblocks (Table 3).

The box plots (Figures 1 and 2) show how well STRAT4 mRNA expression discriminates IHC‐positive from IHC‐negative tumors across specimen types. STRAT4 demonstrated clear separation between IHC‐positive and IHC‐negative tumors for ER across all specimen types, though the separation was less pronounced for PR with core biopsy and cellblock specimens. PR‐positive and PR‐negative cases showed substantial overlap for FNAB‐US specimens. The discriminatory power of STRAT4 between HER2‐positive and HER2‐negative tumors was good for all three specimen types. The cut‐off used by Ng et al. is likely to improve positive and negative predictive values to detect HER2 from cellblock and FNA‐US specimens. The test had low discriminatory power to detect Ki67 for all types of specimens.

FIGURE 1.

Box and whisker plots of STRAT4 delta cycle threshold (dC _t) values for ESR1 (estrogen receptor; ER) in (a–c) and PGR (progesterone receptor; PR) in (d–f) stratified by the ER and PR immunohistochemistry results, respectively, both categorized as negative or positive. STRAT4 ER results for core needle biopsies in (a), cell blocks in (b) and fine needle aspiration in (c) are compared to the manufacturer recommended dC _t cut‐off −1 (dashed red lines) and the modified dC _t cut‐off value of −0.5 (long dashed green lines). STRAT4 PR results for core needle biopsies in (d), cell blocks in (e) and fine needle aspiration biopsy in (f) are compared to the manufacturer recommended dC _t cut‐off −3.5 (dashed red lines) and the modified dC _t cut‐off value of −1.7 (long dashed green lines).

FIGURE 2.

Box and whisker plots of STRAT4 delta cycle threshold (dC _t) values for ERBB2 (human epidermal growth factor receptor 2; HER2) in (a–c) and MKi67 (Antigen Kiel 67; Ki67) in (d–f) stratified by the HER2 immunohistochemistry results, respectively, both categorized as negative or positive. STRAT4 HER2 results for core needle biopsies in (a), cell blocks in (b) and fine needle aspiration in (c) are compared to the manufacturer recommended dC _t cut‐off −1 (dashed red lines) and the modified dC _t cut‐off value of −0.25 (long dashed green lines). STRAT4 Ki67 results for core needle biopsies in (d), cell blocks in (e) and fine needle aspiration biopsy in (f) are compared to the manufacturer recommended dC _t cut‐off −4 (dashed red lines) and the modified dC _t cut‐off value of −2.6 (long dashed green lines).

Agreement in molecular classification of breast cancers between IHC and STRAT4 across different sample types is presented in Table 4. Concordance was 70.8% for core biopsies (κ = 0.60) and 71.7% (κ = 0.61) for cellblocks (71.7%; κ = 0.61), and 57.5% (κ = 0.45) for FNAB‐US samples. The main source of discordance was in the classification of Luminal B tumors, largely driven by discrepancies in Ki‐67 results.

TABLE 4.

Agreement between breast cancer molecular subtypes from IHC and STRAT4 on core biopsy.

Breast cancer	Breast cancer molecular subtype on IHC core biopsy
Molecular subtype	Luminal A	Luminal B	HER2 enriched	Triple negative	Total
On STRAT4 core biopsy
Luminal A	11	5	0	0	16
Luminal B	16	30	3	5	54
HER2 enriched	0	1	36	1	38
Triple negative	0	4	5	20	29
Total	27	40	44	26	137
Agreement	70.8
κ	0.60
On STRAT4 cell blocks
Luminal A	5	1	0	0	6
Luminal B	12	25	3	4	44
HER2 enriched	0	3	26	2	31
Triple negative	1	1	1	15	18
Total	18	30	30	21	99
Agreement	71.7
κ	0.61
On STRAT4 FNA
Luminal A	8	2	0	0	10
Luminal B	15	6	2	1	24
HER2 enriched	0	1	15	5	21
Triple negative	2	1	2	13	18
Total	25	10	19	19	73
Agreement	57.5
κ	0.45

Abbreviations: FNA, fine needle aspirate; HER2, human epidermal growth factor receptor 2; IHC, immunohistochemistry.

We also performed a blinded comparison of IHC on cellblocks and core biopsies (data not shown). We observed moderate to substantial agreement to detect ER (86.7%; κ = 0.72), PR (85.0%; κ = 0.62), and HER2 (75.9%; κ = 0.51). Agreement for detection of Ki‐67 was 78.0% (κ = 0.30).

4. Discussion

Our study observed high concordance between STRAT4 on core biopsy samples and routine IHC for detection of ER. Performance to detect PR and HER2 was variable between the samples, with STRAT4 on FNA‐US samples demonstrating lower concordance with IHC to detect these two markers compared to STRAT4 on core biopsies. Some improvement in agreement was observed using a modified cut‐off. Agreement of the STRAT4 test with routine IHC to detect Ki‐67 was low across all samples, consistent with previous reports.

Performing STRAT4 on manufacturer‐recommended FFPE core biopsy samples remains challenging in LMICs, where core biopsy needles are often unavailable and expensive, procedural costs are high, and histopathology workflows can result in delays exceeding a month [11, 12]. These limitations motivated our evaluation of STRAT4 on samples obtained at FNAB, often the primary mode of tissue diagnosis in the LMICs [13].

Compared to earlier studies, the proportion of samples yielding invalid results was much less in our study. Ng et al. reported 2%–18% of their direct FNA samples and 8% of their cellblock samples to be showing invalid STRAT4 results. Our initial attempts at performing STRAT4 assay on direct FNA samples were challenging in terms of logistics and failed tests because of clotted specimen due to the aspiration of significant blood along with tumor cells. This prompted us to test the unstained smears of FNAB. STRAT4 performed equally well on FNAB‐US and cellblock samples for ER detection, showing almost perfect agreement. Concordance for PR detection was substantial on cellblocks and higher than on FNAB‐US samples; changing dC _t cut‐offs did not alter performance. For HER2, concordance was higher with cellblock samples than FNAB‐US when using manufacturer cut‐offs and improved with modified cut‐offs, particularly for FNAB‐US.

As observed in prior studies, concordance for Ki‐67 was low. A substantial proportion of biopsy samples were positive on STRAT4 but negative on IHC for Ki67, despite using a low threshold (> 15%) on IHC. Both false positives and false negatives on cellblock and FNA samples contributed to low Ki‐67 agreement, which did not improve with modified cut‐offs.

FNAB offers several advantages over core biopsies: it is minimally invasive, cheaper, less technically demanding, and highly accurate for diagnosing malignancy in palpable breast lesions. A meta‐analysis of 46 studies reported pooled sensitivity of 92.7% (95% CI: 92.1–93.3) and specificity of 94.8% (95% CI: 94.3–95.2) for FNAB in detecting breast malignancies in women with palpable lumps. Positive predictive value (PPV) of FNAB to detect breast malignancies exceeds 90% when performed by a trained provider. The GBCI Implementation Framework for national programme managers, policy makers and multisectoral actors has accepted FNAB as a feasible and faster option for tissue diagnostic for resource‐constrained setting [3]. A major limitation of FNA is its inability to recognize in situ carcinoma [14]. However, palpable lumps detected by the women themselves are very unlikely to harbor in situ lesions alone (as seen in our study also). Demonstrating high STRAT4 performance on cellblock or FNA‐US samples to detect ER and moderate to high performance to detect HER2 has important real‐world implications. It enables tailored treatment in resource‐constrained settings using an objective, rapid assay. The Cepheid platform, already widely deployed for HIV, tuberculosis, and increasingly HPV testing in LMICs, does not require highly trained pathologists or technicians.

Preparation of cellblocks from the FNA materials and staining them with immune‐stains permits detection of the biomarkers with reasonable accuracy, as observed in our study; however, cellblock preparation for IHC and slide interpretation pose similar logistical challenges to core biopsies [15]. Using FFPE cellblocks for STRAT4 provides high concordance for ER, PR, and HER2 (especially with the modified HER2 cut‐off) and may be more logistically feasible in LMICs. Importantly, direct analysis of FNAB material bypasses the cost and complexity of cellblock preparation and has been demonstrated in our study that substantial agreement can be achieved for ER. However, detecting HER2 with reasonable accuracy is important for molecular classification of the cancer and management. Agreement to detect this important marker was suboptimal using FNAB samples. Agreement could be improved for FNA‐US specimens (80.8%; κ = 0.57) using the adjusted HER2 cut‐off. Till we have a more reliable detection of HER2 with STRAT4, the current standard of care to detect the marker (IHC/FISH) should be continued before planning for anti‐HER2 therapy.

Detection of Ki‐67 is problematic irrespective of the nature of samples. Even studies comparing Ki‐67 results between core biopsies and surgical specimens have consistently observed lower concordance for this biomarker as compared to the others [16, 17]. The STRAT4 outcomes (on any of the samples) are not suitable to make chemotherapy decisions.

A limitation of STRAT4 is its inability to classify HER2‐low and HER2‐ultralow breast cancers. These categories are currently defined by nuanced immunohistochemical assessment of HER2 protein expression, including staining intensity and pattern, which cannot be directly captured by the quantitative HER2 mRNA measurement provided by STRAT4 [18]. Further studies are needed to determine whether STRAT4‐derived HER2 expression levels can be calibrated against standardized IHC criteria for clinically meaningful classification of HER2‐low and HER2‐ultralow disease.

A key limitation of our study is that the results are preliminary, and larger studies using harmonized FNA processing techniques are needed to establish the clinical utility of STRAT4 on FNA samples. It is to be noted that the discrepancy between a core needle biopsy and FNA for the markers tested may be because of the accompanying blood in FNA, especially when direct unstained smears are used. This does get eliminated to some extent in cellblocks but can persist and can potentially affect PCR‐based amplification. However, the advantage of using direct unstained smears is that this can serve as a rapid “point‐of‐care” test providing molecular information within a few hours enabling a rapid decision on management, a huge advantage in LMIC countries. We have run the test successfully even on smears stored for up to 1 week making it a robust assay that can be provided in higher centers to which smears can be sent. More studies focusing on FNA processing techniques are required to establish the clinical utility of STRAT4 on FNA‐unstained smear samples.

We did not include any cost‐effectiveness analysis. Using a micro‐costing technique, a study in Rwanda demonstrated that using STRAT4 on core biopsy samples saves diagnostic costs by 11% and reduces technicians' and pathologists' time by 61% [19]. Similar studies are required to evaluate cost savings while using STRAT4 on FNA‐derived samples.

5. Conclusion

Our early evaluation study demonstrates mixed results for molecular detection of breast cancer biomarkers using STRAT4 on FNA unstained smears or after preparation of cellblocks. Though the test is performing well to detect ER, the performance needs to improve for other markers. This is feasible by modifying the cut‐off values after collecting data from a higher number of specimens. Our findings are particularly relevant in settings with limited access to histopathology and immunohistochemistry. In a typical secondary‐care workflow in such settings, women presenting with a breast lump may undergo palpation or ultrasound‐guided FNAB with rapid onsite evaluation (ROSE) for immediate triage. In experienced hands, ROSE provides high diagnostic accuracy, with both positive and negative predictive values exceeding 90% [20]. For confirmed malignancies, STRAT4 testing on either cellblock or FNA smears, depending on local resources, can deliver timely biomarker results. This strategy, if observed to be providing satisfactory outcomes, supports decentralization of diagnostic services to secondary‐level facilities, enhancing access and reducing delays—two key strategic objectives of the WHO Global Breast Cancer Initiative [3].

Author Contributions

Partha Basu: investigation, conceptualization, funding acquisition, writing – original draft. Anuradha Talukdar: investigation, writing – original draft, funding acquisition, conceptualization. Poulome Mukherjee: conceptualization, funding acquisition, writing – original draft. Richard Muwonge: conceptualization, funding acquisition. Eric Lucas: conceptualization, writing – original draft, funding acquisition. Monoj Kumar Deka: writing – original draft, investigation. Amanjit Bal: investigation. Upasana Gautam: conceptualization, funding acquisition. Manish Rohilla: conceptualization, writing – original draft. Baneet Bansal: conceptualization, writing – original draft. Veenu Singla: investigation, writing – original draft. Siddhant Khare: writing – original draft. Andre Lopes Carvalho: funding acquisition. Anita Gadgil: conceptualization, writing – original draft. Ravi Kannan: conceptualization, writing – original draft. Radhika Srinivasan: writing – original draft, investigation.

Funding

This work was supported by International Agency for Research on Cancer.

Ethics Statement

Ethical approvals were obtained from the IARC Ethics Committee (IEC 21‐12) and Cachar Cancer Hospital & Research Centre (CCHRC/IRB/012/2021). Written informed consents were obtained from all participants. The study was registered with the Clinical Trials Registry—India (CTRI/2023/03/050522).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Study site‐specific concordance of STRAT4 test outcomes performed on FNAB (unstained smear) samples using two different cut‐offs compared with immunohistochemistry from core biopsies.

Table S2: Study site‐specific concordance of STRAT4 test outcomes performed on core biopsies, cell blocks and fine needle aspitate samples using two different cut‐offs compared with immunohistochemistry from core biopsies.

Basu P., Talukdar A., Mukherjee P., et al., “Automated mRNA‐Based Detection of ER, PR, HER2, and Ki‐67 in Breast Cancer Tissue and Cytology Samples in Resource‐Limited Settings,” Cancer Medicine 15, no. 10 (2026): e72359, 10.1002/cam4.72359.

Data Availability Statement

The data that supports the findings of this study are available from the corresponding author upon reasonable request.