How Do We Identify Patients Who Will Not Receive Chemotherapy After Breast Cancer Surgery?
In this article
Oncotype DX, MammaPrint, Prosigna, EndoPredict and CanAssist Breast do not answer the same question with the same strength. This guide explains which patients’ decisions these tests truly change, which results indicate chemotherapy benefit and which indicate only recurrence risk.
The right patient first: Who were these tests developed for?
The main use of the five tests discussed here is surgically treated hormone receptor-positive, HER2-negative invasive early-stage breast cancer. The central clinical question is whether, in a patient who will definitely receive endocrine therapy, chemotherapy’s additional benefit is large enough to justify its toxicity and quality-of-life costs.
The clinical gray area
In N0 or selected N1 patients, testing changes decisions when pathology neither clearly requires nor clearly rules out chemotherapy.
When the decision is already clear
If a patient will refuse or cannot receive chemotherapy regardless of the result, testing usually satisfies curiosity without changing treatment.
Other biological settings
These tests are not routine solutions for HER2-positive, triple-negative or metastatic disease, pure DCIS or decisions after neoadjuvant treatment.
The critical concept: Recurrence risk and chemotherapy benefit are not the same
“Could the disease return?”
Estimates whether distant recurrence risk during endocrine therapy is low or high. This is the main strength of Prosigna, EndoPredict and CanAssist.
“Will chemotherapy reduce this risk?”
Requires reliable comparisons between patients who received treatment and those who did not. Oncotype DX’s greatest strength is prospective testing of treatment strategies based on its result in TAILORx and RxPONDER.
For example, a patient may have a 12 percent distant recurrence risk, but chemotherapy’s absolute contribution may be 1 percent. Another patient with a similar baseline risk may derive greater benefit. “High risk” therefore does not mean “high chemotherapy benefit.”
How well does clinical appearance reflect biology? Discordance between Ki-67, grade and genomic scores
This may be the guide’s most practical section. In many centers, including in Türkiye, chemotherapy decisions rely on Ki-67, tumor grade and immunohistochemistry-based “Luminal A / Luminal B” classification rather than genomic testing. This is inexpensive and fast, but each of these three measures conflicts with the tumor’s true biology much more often than generally assumed.
How often do Ki-67 and genomic scores disagree?
Contrary to common belief, the relationship between Ki-67 and the Oncotype DX Recurrence Score is weak. In the largest series to date (2.967 patients, Cancers, 2026), the correlation coefficient was only 0,30; agreement between Ki-67 category and genomic risk group was 56,2 percent . In other words, in approximately four out of every ten patients the two methods give different answers—and discordance can run in both directions.
The chart’s most striking line is the first: More than half of patients with high Ki-67 actually have a low genomic score. If chemotherapy had been given based only on Ki-67, many of these patients would have received it unnecessarily. The reverse is more reliable: when Ki-67 is truly low (≤%5), the probability of a low genomic score exceeds 90 percent. In practice, however, only a small minority fall at these clear extremes; most are in the gray zone where Ki-67 says nothing.
An analysis of node-positive patients (298 patients) showed bidirectional discordance: among genomically high-risk tumors (RS 26–100), 29 percent had Ki-67 below %20; among genomically low-risk tumors (RS 0–25), 25 percent had Ki-67 above %20. Discordance decreases markedly only at very high scores (RS 31 and above).
The pattern is similar with other tests: Ki-67 agreement with MammaPrint is 67,1 percent (kappa 0,35), while Ki-67 agreement with EndoPredict is 59,4 percent, not even showing statistically significant agreement (kappa 0,168). Compared with Prosigna, Ki-67-based treatment decisions matched ROR-based decisions only 47 percent of the time. This last study involved only 55 patients and should not alone be considered strong evidence.
Why so much disagreement? The Ki-67 measurement problem
Even after extensive standardization, when Ki-67 was in the critical 10–20 percent band, among 26 cases only 1 case achieved agreement across all laboratories within the same range. The International Ki-67 Working Group therefore recommended in 2021 that Ki-67 be used only at the extremes of %5 or lower (reliably low) and %30 or higher (reliably high) , treating the intervening band as a “gray zone.” In real-world series, 60-80 percent of patients fall precisely in that zone.
How reliable is “Luminal A / Luminal B” classification?
Immunohistochemical Luminal A / Luminal B classification (based on ER, PR, HER2 and Ki-67), compared with true molecular subtyping by PAM50, disagrees approximately one-third of the time (in a 607-patient series, discordance was 38,4 percent). Notably, tumors called “Luminal B” by immunohistochemistry were also Luminal B by PAM50 in only 53 percent—almost a coin toss. In tumors with Ki-67 between 10-30 percent, classification becomes entirely uncertain.
The practical meaning for patients is that “Luminal B” in a pathology report is not a molecular diagnosis but an approximate estimate; its most frequent error is to label a truly Luminal A tumor as Luminal B, leading to unnecessary chemotherapy.
Does tumor grade reflect biology?
In other words, approximately one-quarter to one-third of grade 3 tumors are genomically low risk —and chemotherapy has shown no measurable survival contribution in these patients. In SEER data, node-negative small grade 3 tumors with low genomic scores had 5-year breast cancer-specific survival above 99 percent. The reverse is much rarer: in grade 1 tumors, high genomic scores occur in about 3-9 percent, and in some series not at all.
The 2026 evidence map for five tests
The map below compares prospective randomized data, prediction of chemotherapy benefit, late recurrence assessment and current guideline support, rather than marketing claims.
* Guidelines and payers do not accept the same tests in the same ways. For example, ASCO and NICE differ on MammaPrint and node-positive use. As of August 2026, Prosigna OPTIMA data are a conference presentation; peer-reviewed full publication and guideline integration are awaited.
Oncotype DX: The test most directly assessing chemotherapy decisions
21 genes | Recurrence Score 0–100 | Central laboratory
Oncotype DX measures activity of 21 genes in tumor tissue, giving a Recurrence Score from 0 to 100. It helps assess distant recurrence risk and expected chemotherapy benefit in certain groups. Its greatest advantage is two major randomized trials that changed practice: TAILORx in node-negative disease and RxPONDER with 1–3 positive nodes.
The 12-year TAILORx message
In node-negative HR-positive/HER2-negative disease, adding chemotherapy for the overall RS 11–25 group provided no significant overall advantage during long follow-up. At 12 years, freedom from distant recurrence was 92,6 percent and 92,8 percent, with overall survival of 89,8 percent in both randomized arms. However, a small benefit signal persisted in some patients aged 50 or younger, particularly RS 21–25 or RS 16–20 combined with high clinical risk.
In the RS 0–10 group, 12-year freedom from distant recurrence with endocrine therapy alone was 93,2 percent. Chemotherapy is standard for RS 26–100, but this high-score arm was not randomized in TAILORx. Benefit at high scores is based on combined interpretation of the older NSABP B20 comparison and modern cohorts.
RxPONDER: Menopausal status is decisive in node-positive disease
RxPONDER assessed patients with 1–3 positive nodes and RS 0–25, totaling 5.018. In postmenopausal patients, adding chemotherapy did not improve 5-year invasive disease-free survival. In premenopausal patients, the 5-year result increased from 89,0 percent to 93,9 percent.
Current data showing that small tumors are not always biologically low risk are discussed below in the small-tumor section and our analysis of small tumors and Oncotype DX in detail.
MammaPrint: Can high clinical risk coexist with low genomic risk?
70 genes | Low / high, plus ultralow and High 1–2 classes | Central analysis
MammaPrint uses a 70-gene signature to classify tumors as genomically low or high risk. Its key study is MINDACT, which tested the ability to guide chemotherapy decisions when clinical and genomic risks disagree.
MINDACT itself provides the largest-scale evidence of the discordance described earlier: among its 6.693 participants, 23,2 percent were clinically high risk but genomically low risk, 8,8 percent were clinically low risk but genomically high risk. Thus, clinical assessment and tumor biology gave different messages in approximately one-third of patients.
In the clinically high/genomically low-risk group, eight-year distant metastasis-free survival was 92,0 percent with chemotherapy and 89,4 percent without it, an absolute difference of 2,6 points. This does not mean “nobody with a low MammaPrint score benefits from chemotherapy.” More accurately, selected patients may omit chemotherapy while accepting a small potential benefit forgone.
Age interaction was exploratory, not a prespecified primary endpoint. The approximately 5-point difference in younger patients may also reflect chemotherapy-induced ovarian suppression, as in TAILORx and RxPONDER.
Ultralow risk
In MINDACT’s ultralow genomic-risk patients, 8-year breast cancer-specific survival was 99,6 percent and distant metastasis-free interval 97,0 percent. This group matters for de-escalation research, but the test alone does not instruct patients to “stop endocrine therapy.”
FLEX registry
In a 2025 real-world analysis, estimated 5-year chemotherapy benefit was below 1 percent in the ultralow group, 1,7 in Low, in High 1 it was 5,6, and in High 2 it was 10,9 points. This was not randomized; it generates a strong hypothesis but is not on the same evidence level as MINDACT.
Prosigna: Boundaries changing with OPTIMA 2026
PAM50 / 50 genes | ROR 0–100 | Molecular subtype + proliferation + tumor size
Prosigna generates a ROR (Risk of Recurrence) score from the PAM50 signature and describes tumors by similarity to Luminal A, Luminal B, HER2-enriched or basal-like biological subtypes. Its historical strength is predicting 10-year and late distant recurrence risk, especially in postmenopausal patients.
The phase III OPTIMA study reported in 2026 is a milestone for this test. Patients aged 40 and older with ER-positive/HER2-negative disease and clinical indications for chemotherapy, totaling 4.429, were randomized to standard chemotherapy + endocrine therapy or a Prosigna-guided strategy. In the test arm, those with ROR 60 or lower received endocrine therapy alone, omitting chemotherapy in 68 percent of that arm.
The test-guided strategy was noninferior to standard treatment within the prespecified statistical boundary (adjusted HR 1,03; %90 CI 0,85–1,25; the prespecified 3 percent noninferiority margin was met). Results were consistent in premenopausal patients and the 1–3 positive-node subgroup; approximately three-quarters of the per-protocol population already had 1–3 positive nodes. Mandatory ovarian suppression for premenopausal test-arm patients not receiving chemotherapy importantly distinguishes OPTIMA from earlier studies.
Another strength of Prosigna is late recurrence. ABCSG-8 data showed distant recurrence risk during years 5–15 of approximately 2,4 percent in low ROR and 17,5 percent in high ROR. This can inform extended endocrine therapy discussions but does not directly predict the benefit of extension—for that distinction see the Breast Cancer Index section .
EndoPredict: EPclin combines molecular biology and clinical burden
12 genes | EP molecular score + tumor size + nodal status = EPclin | Available in local laboratories
EndoPredict’s clinically used EPclin score combines gene expression, tumor size and lymph node status, incorporating tumor biology and anatomical disease burden in one model. EPclin 3,3 or lower is generally low risk, and higher values high risk.
Its strength, particularly in ER-positive/HER2-negative patients receiving endocrine therapy, is distinguishing distant recurrence risk over 10–15 years. Its weakness is the absence of a randomized test-strategy trial like TAILORx, RxPONDER or MINDACT assessing chemotherapy benefit.
This difference supports potential chemotherapy benefit, but because treatment was not randomized, “20 points of chemotherapy benefit” cannot be inferred from the chart. Another analysis combining cohorts suggested greater benefit at high EPclin scores, but again was not direct randomized evidence.
CanAssist Breast: An accessible protein test emerging from South Asia
5 immunohistochemical proteins + 3 clinical features | Machine learning | 5-year distant recurrence risk
Unlike the other four tests, CanAssist Breast is not a gene-expression test. It combines CD44, ABCC4, ABCC11, N-cadherin and pan-cadherin with tumor size, grade and nodal status in a machine-learning model. Its development in India and aim of applicability in lower-resource health systems are important features.
Scores range from 0–100; 15,5 or below is low risk and above that high risk. In a real-world series published in 2025 involving 5.926 patients, 72 percent were low risk and 28 percent high risk.
Extensive use does not mean extensive outcome evidence. In the 2025 analysis, among 583 patients completing five years, follow-up data existed for only 305; mature outcome analysis was planned for 2029. There is no randomized chemotherapy-decision trial, and major international guidelines do not place it on equal footing with Oncotype DX, MammaPrint, Prosigna or EndoPredict.
Oncotype, MammaPrint, Prosigna, EndoPredict and CanAssist: One table
| Feature | Oncotype DX | MammaPrint | Prosigna | EndoPredict | CanAssist Breast |
|---|---|---|---|---|---|
| Technology | 21 genes, RT-PCR | 70 genes, microarray | 50 genes, nCounter/PAM50 | 12 genes, RT-PCR; EPclin adds clinical data | 5 proteins, IHC + machine learning |
| Report | RS 0–100; interpreted by age/nodal context | UltraLow, Low, High 1, High 2 | ROR 0–100 and intrinsic subtype | EPclin low / high | ≤15,5 low; >15,5 high |
| Main question | Recurrence risk and chemotherapy contribution | Can chemotherapy be omitted at high clinical risk? | 10-year and late recurrence; treatment strategy with OPTIMA | 10–15-year distant recurrence | 5-year distant recurrence risk |
| Pivotal study | TAILORx, RxPONDER | MINDACT | OPTIMA 2026; ABCSG-8 | ABCSG-6/8, TransATAC and real-world cohorts | Retrospective validations and a registry of 5.926 tests |
| Randomized clinical benefit | Yes, strong | Yes, strong | Yes; 2026 conference data, short follow-up | No direct evidence | None |
| Integration with clinical data | Through a separate tool: RSClin / RSClin N+ | No combined score; clinical risk calculated separately | ROR already includes tumor size | EPclin is integrated by design | The model includes clinical variables |
| Where performed | Central laboratory only (United States) | Central shipment (United States/Netherlands) | Can be performed in a local laboratory | Can be performed in a local laboratory | Central (India) |
| N0 | Most established option | Particularly clinically high/genomically low | Established postmenopausally; OPTIMA broader | Appropriate postmenopausally | Developing data |
| 1–3 nodes | Strong in postmenopausal RS 0–25 | MINDACT subgroup; small absolute difference | OPTIMA and earlier prognostic data | Selected postmenopausal patients | Limited |
| Premenopausal | Usable in N0; low N1 score does not exclude chemo | Benefit signal in younger patients and OFS uncertainty | Promising in OPTIMA at age ≥40 with mandatory OFS | Increasing data, weak decision evidence | Not yet sufficient |
| Main limitation | Score cannot be interpreted apart from clinical context | Binary risk category is not equivalent to chemotherapy benefit | Awaiting full OPTIMA publication and longer follow-up | Indirect predictive evidence | No mature outcomes or randomized benefit data |
RSClin: A tool combining score and clinical information
The principle repeated throughout this guide—“the score is not read alone”—has taken the form of a concrete calculator since 2021. RSClin combines the Oncotype DX Recurrence Score with tumor grade, tumor size and age to produce two personalized figures rather than a categorical threshold: 10-year distant recurrence risk with endocrine therapy and chemotherapy’s absolute contribution for that patient.
Why is it better than the score alone?
In a development set of 10.004 patients (NSABP B-14, B-20 and TAILORx), RSClin was statistically more accurate than clinical features alone or Recurrence Score alone. Independent validation used the 1.098-patient Israeli Clalit real-world cohort, with agreement between predicted and observed risk of 0,96 .
The same score, very different benefit
Its most instructive finding: in a patient aged 55 with a 1,5 cm intermediate-grade tumor, chemotherapy contribution across RS 11–50 ranged from 0 to 15 percent ; for a 2,5 cm high-grade tumor, the same score range corresponded to 1 to 33 percent benefit. In other words, the same Recurrence Score may imply completely different chemotherapy decisions in different patients.
Published in 2024, the RSClin N+ version expanded to node-positive patients using SWOG S8814 and RxPONDER data (5.283 patients), adding the number of positive nodes. In postmenopausal patients, absolute chemotherapy benefit across scores ranges from 0,1 to 21,5 percent . But an important limitation exists: no significant interaction between score and chemotherapy benefit was found in premenopausal patients —RSClin N+ therefore does not identify young node-positive patients who can avoid chemotherapy. This is consistent with RxPONDER itself.
Other tests build integration into their design: EndoPredict’s EPclin already includes tumor size and nodal status; Prosigna’s ROR accounts for tumor size. MammaPrint has no validated combined score—the MINDACT approach is to place genomic risk and separately calculated clinical risk side by side, not merge them into one number.
A practical decision algorithm for 2026
These pathways organize clinical discussions; they do not replace a tumor board.
Gate 1: Before ordering a genomic test
2. In node-negative patients
3. With one–three positive nodes
4. With four or more positive nodes
Small tumors: Does size determine biology?
For node-negative tumors smaller than one centimeter, the usual assumption is: “Small means low risk; neither testing nor chemotherapy is needed.” A very large study published in 2026 directly tested this assumption—and the situation is less simple than it seems.
In other words, approximately one in every nine to ten tumors ≤1 cm is genomically high risk. This is not a negligible proportion. In the same study, among these high-score patients, 10-year overall survival comparing those receiving chemotherapy with those not receiving it was 90,4 percent versus 81,9 percent ; the propensity-score-adjusted hazard ratio was 0,71. The difference was most apparent in the 5–10 mm (T1b) group.
Absolute numbers help provide balance. SEER data show 5-year breast cancer-specific mortality in node-negative disease of 0,4 percent at low scores and 4,4 percent at high scores—a tenfold relative difference from a small absolute baseline. In NCDB, the number needing treatment to prevent one death over 5 years was approximately 44.
T1b (5–10 mm), grade 3, LVI present
This group has the highest probability of a high score; among ≤1 cm grade 3 tumors, 32,1 percent have high scores. The result can genuinely change the decision.
T1b, grade 2, no LVI
Testing is possible, but absolute benefit remains small even with a high score, and MINDACT does not support chemotherapy benefit. Patient preference should inform the decision.
T1a / T1mi, grade 1
Absolute risk is so low that even a high score rarely changes the arithmetic. Current guidelines do not require testing below 5 mm.
We also discussed this in our article on small breast tumors and Oncotype DX .
Special situations complicating decisions
Lobular cancer
Although included in studies, invasive lobular cancer subgroups are smaller and have a different late-recurrence pattern. Low early risk does not mean zero lifetime risk.
Multifocal / multicentric disease
If foci differ in histological grade or receptors, choose the most representative or aggressive focus. Multiple tests are not routinely necessary; pathology–oncology communication is essential.
Male breast cancer
Real-world data help, but large randomized studies predominantly include women. Results must be interpreted acknowledging this lack of direct evidence.
Older age and comorbidity
High genomic risk does not mean chemotherapy is tolerable. Geriatric assessment, life expectancy and absolute benefit are decisive.
After neoadjuvant treatment
The strongest evidence comes from postoperative samples before systemic therapy. Routine use in treated residual tumors is not established.
Very low PR or ER
Tumors with borderline ER positivity (1–10 percent) may resemble triple-negative disease biologically. Genomic test evidence is weak here; decisions integrate clinical and pathological findings.
Six common mistakes
Breast Cancer Index: A different question after five years
All five tests in this guide answer the question at diagnosis : “Should I receive chemotherapy now?” But patients frequently ask a second, entirely different question: “I have completed five years of endocrine therapy and remain disease-free. Should I continue for another five years?”
Breast Cancer Index (BCI) combines the HOXB13/IL17BR gene ratio with the molecular grade index and is the only test making a predictive claim for this second question. Analyses of archived tissue from three studies showed a consistent pattern:
| Study | Extended treatment | Benefit in BCI(H/I)-high group | In BCI(H/I)-low group |
|---|---|---|---|
| MA.17 | Letrozole after 5 years | Marked recurrence-risk reduction; absolute benefit ~%16,5 | No benefit |
| Trans-aTTom (node-positive) | 10 years vs 5 years tamoxifen | Absolute benefit %9,7 (HR 0,33) | No benefit (HR 1,11) |
| IDEAL | 2,5 years vs 5 years extended letrozole | Absolute benefit ~%9,8 | No benefit |
IDEAL has a particularly striking finding that reaffirms this guide’s central theme: approximately half of clinically high-risk patients were in the BCI group not benefiting from extension, while about half of clinically low-risk patients were in the benefiting group. Again, clinical appearance and biology diverge.
Guideline positioning is measured: ASCO’s 2022 biomarker guideline states that after completing 5 years of endocrine therapy, node-negative or 1–3 positive-node patients “may be offered” BCI (moderate-quality evidence, moderate-strength recommendation); NCCN lists it as category 2A for extended endocrine therapy decisions and states it is the only test with this indication.
ctDNA and residual disease: Exciting, but not yet a decision tool
Detecting microscopic postoperative residual disease by measuring circulating tumor DNA (ctDNA) attracts great interest in breast cancer, as in colorectal cancer. Patients increasingly ask: “Could a blood test tell us whether chemotherapy is needed?”
Studies show both the technology’s power and its fundamental problem:
- c-TRAK TN (early triple-negative disease): among ctDNA-positive patients, 72 percent already had disseminated disease on imaging; testing often acted as late confirmation rather than “early warning.” No treated patient achieved sustained ctDNA clearance.
- LEADER (HR+/HER2−): 33 percent of ctDNA-positive patients already had visible metastases.
- TRAK-ER: at first detection, disease had already recurred in 42,9 percent; additionally, test-specific panel design failed in 16,1 percent —a useful reminder that molecular tests can also fail technically.
- ZEST: finding patients with truly molecular-only recurrence (clear imaging) was so difficult that the study closed early because recruitment was insufficient.
- DARE: ctDNA negativity had a very high negative predictive value (at month 6, 100 percent). However, among 494 screened patients, only 40 could be randomized, and outcome data are not yet available.
Can it guide adjuvant chemotherapy decisions? Absolutely not. In HR+/HER2-negative early breast cancer, there is no evidence supporting decisions to give or omit chemotherapy based on ctDNA, and no test has been validated for this purpose. Outside clinical trials, patients purchasing it are paying for information that cannot guide action.
The practical side: Tissue, timing and logistics
Test choice depends on more than evidence strength. Remaining tumor tissue, turnaround time and testing location influence real-world decisions at least as much as evidence.
| Test | Where performed | Tissue requirement | Laboratory time |
|---|---|---|---|
| Oncotype DX | Central laboratory only (United States); shipment from Türkiye | Paraffin block or 15 unstained sections; microinvasive foci not accepted | 7–10 days + shipping |
| MammaPrint | Central shipment (United States/Netherlands); NGS version can be performed locally | Paraffin tissue, at least %30 tumor cells; lymph node material not validated | Usually up to 10 days + shipping |
| Prosigna | In a local laboratory, can be performed | At least 4 mm² tumor area and %10 viable tumor; macrodissection mandatory | ~3 days |
| EndoPredict | In a local laboratory, can be performed (kit-based) | At least %30 tumor cells; fixation time critical | ~3–5 days |
| Breast Cancer Index | Central (United States) | Block or 3–4 sections; invasive tumor area ≥2 mm; core biopsy explicitly accepted | ~10 days + shipping |
Central or local? A critical difference for Türkiye
Prosigna and EndoPredict can be performed in a laboratory in Türkiye, with results in 5-7 days. Oncotype DX, MammaPrint and BCI are sent abroad; adding shipping, customs and paperwork to laboratory time may extend the adjuvant decision to 2 weeks . Blocks or sections also physically leave the country, a separate risk when tissue is limited and further testing may be needed.
But the strongest evidence belongs to send-out tests
The honest balance is this: the largest randomized studies, such as TAILORx, RxPONDER and MINDACT, concern the two centrally shipped tests (Oncotype DX and MammaPrint); RSClin also exists only for Oncotype. Speed and logistical convenience stand against evidence strength here.
Access, reimbursement and cost in Türkiye
Scientific value is not the same as practical accessibility in Türkiye. The table below uses information available as of August 2026. Because this field changes quickly, always confirm the current situation with your treating center.
| Test | Access in Türkiye | Note |
|---|---|---|
| Oncotype DX | Available—through shipment abroad (e.g. Genekor) | All samples processed at the central laboratory in the United States |
| Prosigna | Available— performed within Türkiye | The nCounter platform was introduced in 2018 through METU’s Informatics Institute; currently offered by private laboratories |
| MammaPrint | Available—through shipment abroad | Turkish instructions for use available |
| Breast Cancer Index | Available—through shipment abroad | For extended endocrine therapy decisions |
| EndoPredict | Available | |
| CanAssist Breast | Available | Mainly in India and neighboring countries |
Cost: A prospective cost-effectiveness study across 10 centers and 165 patients in Türkiye (European Journal of Breast Health, 2019), using an Oncotype DX cost of 5.141 dollars, calculated the cost per quality-adjusted life year gained as 7.207,90 dollars and recommended inclusion in standard care in our country. This recommendation has not entered SUT coverage in the approximately seven intervening years. A 2025 clinician source reports approximately 3.000 euros . No current reliable prices were found for other tests; exact figures require direct consultation with the relevant laboratory.
If choosing a test, which clinical question should guide selection?
| Clinical question | Leading test / approach | Why? |
|---|---|---|
| Does chemotherapy benefit an N0 patient? | Oncotype DX | The most direct randomized evidence through TAILORx. |
| Postmenopausal, 1–3 positive nodes and RS 0–25 | Oncotype DX | RxPONDER showed no additional chemotherapy benefit. |
| If clinical risk is high but genomic risk low, can chemotherapy be omitted? | MammaPrint | MINDACT directly tested this discordance. |
| Age ≥40, high clinical risk; test-guided strategy with greater nodal burden | Prosigna, in the OPTIMA context | 2026 randomized noninferiority data; evidence still maturing. |
| Combining the score with age, tumor size and grade to see the absolute chemotherapy contribution | RSClin / RSClin N+ | Provides an individualized percentage rather than a categorical threshold; validated in an independent cohort. |
| Distinguishing 10–15-year late recurrence risk | Prosigna or EndoPredict | Strong prognostic validations with long follow-up. |
| Should endocrine therapy be extended after five years? | Breast Cancer Index | The only test making a predictive claim for this question; guideline level “may be offered.” |
| Identifying ultralow biological risk | MammaPrint UltraLow | Excellent long-term prognosis; valuable for de-escalation research. |
| When results are needed quickly or tissue is scarce | Prosigna (local, ~3 days, 4 mm²) | The most logistically tolerant option; evidence not as broad as Oncotype. |
| A more accessible regional protein-based option | CanAssist Breast | Promising and pragmatic, but evidence remains insufficient to exclude chemotherapy. |
The future: Predicting scores from pathology slides and smarter decision models
Deep-learning studies published in 2026 approached predicting whether Oncotype DX scores are high from routine H&E pathology images. Models developed in TAILORx and external cohorts performed impressively but do not replace certified tests today. The nearest potential use may be confidently predicting low risk to reduce the number needing expensive molecular tests—particularly meaningful in countries such as Türkiye without reimbursement. For details, see Oncotype prediction from pathology slides in our article.
Another critical future question is how much younger patients’ benefit comes from true cytotoxic chemotherapy versus ovarian suppression. The 2026 EBCTCG meta-analysis strengthened evidence that ovarian suppression reduces recurrence in appropriate premenopausal women. The randomized OFSET study directly compares chemotherapy + optimal endocrine therapy with optimal endocrine therapy alone in low/intermediate genomic risk. Its answer is the single finding that could change all premenopausal recommendations in this guide.
For breast cancer stages and all treatment components, consult our guide to breast cancer symptoms, stages and treatment; for the basic logic of genomic testing, our chemotherapy and genomic testing guide (Türkçe) is available.
The aim is to reduce inappropriate chemotherapy, not simply chemotherapy
Identifying patients who can safely omit postoperative chemotherapy requires much more than reading one number. The strongest decision combines pathology, anatomical risk, menopausal status, an optimal endocrine therapy plan, overall health and the evidence for the right test.
The most recurrent finding in this guide is: a tumor’s microscopic appearance conflicts with its biology much more often than assumed. More than half of patients with high Ki-67 have low genomic scores; one-quarter to one-third of grade 3 tumors are genomically low risk; and approximately one-third in MINDACT had discordant clinical and genomic risks. These conflicts are not errors but the very reason these tests exist.
- Oncotype DX is the strongest test directly positioning chemotherapy benefit; combined with RSClin, it gives personalized absolute benefit.
- MammaPrint is strong in clinically high/genomically low discordance and ultralow risk.
- Prosigna is strong in late recurrence; OPTIMA 2026 may substantially expand its treatment-decision role, and local testing in Türkiye is a practical advantage.
- EndoPredict is a strong long-term prognostic tool combining clinical burden and genomic biology; chemotherapy-benefit evidence is indirect.
- CanAssist Breast is an important India-origin accessibility initiative, but not currently evidence-equivalent to the other four tests.
- Breast Cancer Index answers a separate question: whether to extend endocrine therapy after five years.
- ctDNA is exciting but cannot currently guide adjuvant chemotherapy decisions.
The best question is not “Which test is best?” but “Which evidence from which test can resolve which uncertainty in this patient?”
Main scientific references
- Sparano JA et al. Clinical and Genomic Risk for Late Breast Cancer Recurrence and Survival. NEJM Evidence, 2024. TAILORx 12-year update.
- Kalinsky K et al. 21-Gene Assay to Inform Chemotherapy Benefit in Node-Positive Breast Cancer. NEJM, 2021. RxPONDER.
- Piccart M et al. 70-gene signature as an aid for treatment decisions in early breast cancer: updated MINDACT results. Lancet Oncology, 2021. MINDACT long-term follow-up.
- Lopes Cardozo JMN et al. Outcome of Patients With an Ultralow-Risk 70-Gene Signature in the MINDACT Trial. JCO, 2022. MammaPrint UltraLow.
- Earl HM et al. OPTIMA phase III randomized trial. ASCO 2026, JCO 2026;44(16_suppl):500. OPTIMA abstract.
- Dowsett M et al. Prediction of late distant recurrence after 5 years of endocrine treatment: PAM50 ROR. JCO, 2013/2014. Prosigna late recurrence data.
- Long-Term Prospective Outcome Data Using EndoPredict. Breast Care, 2024. EndoPredict real-world cohort.
- Real-world clinical application of CanAssist Breast among 5,926 patients. Scientific Reports, 2025. CanAssist Breast real-world data.
- Sparano JA et al. Development and Validation of a Tool Integrating the 21-Gene Recurrence Score and Clinicopathologic Features (RSClin). JCO, 2021;39:557-564. RSClin development and validation.
- RSClin N+: Combining the 21-gene Recurrence Score and clinicopathological features in node-positive disease (based on SWOG S8814 and RxPONDER). JCO, 2024. RSClin N+.
- House A, Stephens JA, Wesolowski R et al. Impact of Oncotype DX risk categorization and receipt of chemotherapy on survival outcomes among patients with small node-negative HR+ breast cancer. The Oncologist, 2026;31(8). NCDB analysis of small tumors.
- Petkov VI et al. Breast-cancer-specific mortality in patients treated based on the 21-gene assay: a SEER population-based study. npj Breast Cancer, 2016;2:16017. SEER mortality data.
- Analysis of agreement and discordance between Ki-67 and the 21-gene Recurrence Score. Cancers, 2026;18(11):1731. Correlation study of 2.967 patients.
- Comparison of International Ki-67 Working Group thresholds with the 21-gene test. Breast Cancer Research and Treatment, 2024;203. Kaiser Permanente cohort.
- Nielsen TO et al. Assessment of Ki67 in Breast Cancer: Updated Recommendations From the International Ki67 in Breast Cancer Working Group. JNCI, 2021;113:808. Ki-67 recommendations.
- Polley MC et al. An international Ki67 reproducibility study. JNCI, 2013;105:1897 ve Modern Pathology, 2015. Interlaboratory variability.
- Grade 3 ER-positive disease: 21-gene score and chemotherapy benefit, National Cancer Database analysis. JCO Precision Oncology, 2020. Discordance between grade and genomic risk.
- Agreement between PAM50 molecular subtype and immunohistochemical subtype. Cancer Research and Treatment, 2019. Subtype discordance.
- Sgroi DC et al. Prediction of late disease recurrence and extended adjuvant letrozole benefit by the HOXB13/IL17BR biomarker (MA.17). JNCI, 2013;105:1036. BCI — MA.17.
- Bartlett JMS et al. Breast Cancer Index Is a Predictive Biomarker of Treatment Benefit and Outcome from Extended Tamoxifen Therapy: Final Analysis of Trans-aTTom. Clinical Cancer Research, 2022;28:1871. Trans-aTTom.
- Noordhoek I et al. Breast Cancer Index Predicts Extended Endocrine Benefit (IDEAL). Clinical Cancer Research, 2021;27:311. IDEAL.
- Andre F et al. Biomarkers for Adjuvant Endocrine and Chemotherapy in Early-Stage Breast Cancer: ASCO Guideline Update. JCO, 2022. ASCO biomarker guideline.
- NICE. Tumour profiling tests to guide adjuvant chemotherapy decisions in early breast cancer. 2024. NICE HTG719.
- Karadeniz Çakmak G et al. Cost-Effectiveness of Gene Expression Profiling in Patients with Early-Stage Breast Cancer in a Middle-Income Country: Results of a Prospective Multicenter Study (Türkiye). European Journal of Breast Health, 2019;15(3). Türkiye cost-effectiveness study.
- Turnbull AK et al. and ESMO Breast Cancer 2026. The role of ctDNA surveillance in early-stage breast cancer. ESMO 2026 — ctDNA surveillance.
- Early Breast Cancer Trialists’ Collaborative Group. Ovarian function suppression in premenopausal ER-positive breast cancer. Lancet, 2026. Patient-level meta-analysis.
Medical note: This content provides general information, not personalized treatment recommendations. Test availability, licensing, reimbursement conditions and guideline positioning may vary by country and date. Decisions require multidisciplinary assessment involving breast surgery, pathology, medical oncology, radiation oncology and, when necessary, genetics.
Sources
Impact of Oncotype DX Recurrence Score on Treatment Decisions: Results of a Prospective Multicenter Study in Turkey.
Cureus. 2016 Mar 8;8(3):e522
2. Siegel RL, Miller KD, Jemal A.
Cancer Statistics, 2017.
CA Cancer J Clin. 2017 Jan 5.
3. Sparano JA, Gray RJ, Makower DF, et al.
Prospective Validation of a 21-Gene Expression Assay in Breast Cancer.
N Engl J Med. 2015;373(21):2005.
4. Parker JS, Mullins M, Cheang MC, et al.
Supervised risk predictor of breast cancer based on intrinsic subtypes.
J Clin Oncol. 2009;27(8):1160.
5. Muaiad Kittaneh, Sunil Badve, Humberto Caldera, Lee Schwartzberg, Tiffany Traina, Charles Vogel et al.
Case-Based Review and Clinical Guidance on the Use of Genomic Assays for Early-Stage Breast Cancer: Breast Cancer Therapy Expert Group (BCTEG).
Clinical Breast Cancer. 08 Jan 2020.
Was this article helpful?
Your feedback helps us improve our content.
Reader comments
Share your comment
Your name and comment are published only after administrator approval. Your email address is never shown to visitors.
Be the first to comment.