The August 2026 issue of CAP TODAY headlines Homologous recombination, issues and dilemmas. We put that together with a 2022 review in The Oncologist and a 2042 review in CAP TODAY.
##
The central lesson from CAP Today in 2026
Karen Titus’s 2026 article is best understood as an update on a problem that the 2022 Friends of Cancer Research paper hoped the field would solve but has not yet solved: HRD is clinically important, but it is not a single analyte with a standardized measurement system.
“HRD testing” can mean at least three different things:
Finding a possible cause of deficient repair, such as a pathogenic BRCA1/2 alteration.
Measuring the accumulated consequences of deficient repair—LOH, telomeric allelic imbalance, large-scale state transitions, and related genomic scars.
Determining whether the tumor is functionally HR-deficient now, at the moment treatment is being considered.
Current NGS-based clinical testing does the first two reasonably well under favorable conditions. It does not necessarily answer the third. Titus’s deeper message is therefore not merely that HRD testing is technically difficult. It is that the laboratory is being asked to turn several biologically related but nonidentical measurements into a binary clinical answer.
The article the file labels as “2025” was actually published by CAP on May 29, 2024. The January 1, 1970 date on the article page is a website artifact; CAP’s Precision Medicine article index gives the correct date.
A useful teaching framework: cause, scar, and current function
| Clinical question | What is measured | Principal strength | Principal limitation |
|---|---|---|---|
| What may have caused HRD? | Germline or somatic BRCA1/2 and other HRR-gene alterations; potentially methylation | Identifies a specific biological lesion; BRCA findings may have hereditary implications | A mutation in an HRR-associated gene does not invariably establish an HRD phenotype; methylation, large deletions, and some copy-number losses may be missed |
| Has the tumor experienced HR failure? | LOH, TAI, LST, or composite genomic-instability scores | Captures HRD beyond BRCA and other readily identifiable causal variants | Genomic scars are historical, persistent, continuous, spatially heterogeneous, and assay-dependent |
| Is the tumor HR-deficient today? | Potential functional assays such as RAD51 foci | Conceptually closest to the current drug-sensitive phenotype | Not yet established as a practical, broadly validated routine assay |
| What broader biology surrounds the result? | Comprehensive genomic profiling: CCNE1, RB1, HRR genes, resistance alterations, MSI/TMB, and other biomarkers | Can help interpret borderline or biologically contradictory results | Much of this contextual interpretation remains exploratory rather than part of a validated HRD classifier |
This distinction is already explicit in the 2022 Oncologist paper, which separates potential “causes” of HRD from their genomic “consequences.” Titus makes the distinction more clinically vivid: the scar is evidence that homologous recombination failed, but the scar itself neither causes PARP sensitivity nor proves that the repair defect is still present.
The principal problems identified in Titus 2026
1. HRD is not one measurement
BRCA1/2 alterations remain the strongest and best-understood causal biomarkers. But BRCA testing alone misses tumors that have acquired an HRD phenotype through other mechanisms. Conversely, merely finding an alteration in PALB2, CHEK2, ATM, BRIP1, RAD51C, RAD51D, or another nominal HRR gene does not mean that every alteration has the same penetrance, produces biallelic loss, creates the same degree of genomic instability, or predicts the same treatment benefit.
Titus quotes Kyle Strickland bluntly: “just looking at the genes that are altered is not a good way to evaluate HRD.” That is an important correction to the simplistic equation:
HRR-panel mutation = HRD-positive.
Gene testing and genomic-scar testing provide complementary information. Neither automatically subsumes the other.
2. The clinical-trial evidence does not use a uniform biomarker
The major PARP-inhibitor trials differ simultaneously in drug, treatment setting, combination therapy, enrolled population, biomarker definition, and numerical cutoff:
SOLO1 was overwhelmingly a BRCA-mutated population.
PRIMA defined HRD through a deleterious BRCA alteration or a genomic-instability score of at least 42.
PAOLA-1 also used 42, but studied olaparib combined with bevacizumab.
VELIA used a cutoff of 33.
Thus, 33 versus 42 is not merely an analytical disagreement that laboratories can settle by choosing the statistically “best” threshold. The cutoffs are embedded in particular trial designs and treatment claims. A result cannot be interpreted independently of the assay, cancer population, therapeutic regimen, and validation study that produced the cutoff.
This was already one of the 2022 paper’s central concerns: different assays and thresholds can produce different HR-status calls and consequently different treatment decisions. Titus shows that this remains a live clinical problem in 2026. The CAP Today article summarizes these trial and cutoff differences directly.
3. A continuous, heterogeneous signal is forced into a binary result
Clinicians understandably want “HRD positive” or “HRD negative.” But genomic instability is measured on a continuous scale. The cutoff converts that continuous measurement into a category for a specific clinical purpose; it does not establish a natural biological border.
Titus provides an especially useful example: different regions of the same heterogeneous ovarian carcinoma can fall on opposite sides of the threshold. A tumor tested four times might be classified as deficient three times and proficient once. A score of 41 is analytically reported as negative under a cutoff of 42, but it is not biologically the opposite of a score of 42.
This raises several unresolved questions:
Should there be a formal indeterminate or borderline interval?
Should another tumor block be tested near the cutoff?
How much movement results from true intratumoral heterogeneity versus analytical variation?
Should a borderline score be interpreted alongside BRCA and other genomic findings?
Is a cutoff validated in one therapeutic context transportable to another?
Titus does not supply definitive answers. Her point is that the clean binary result received by the oncologist conceals a substantial gray zone.
4. The specimen can determine the answer
The article gives preanalytics unusual prominence, and rightly so. Ovarian cancer patients may receive neoadjuvant chemotherapy before debulking surgery. Treatment can eradicate or greatly reduce viable tumor, forcing the laboratory back to a small pretreatment biopsy.
The requirements are particularly demanding for copy-number loss. Strickland says that accurate detection of BRCA1/2 copy-number losses may require approximately 50 percent tumor content on the slide. Small biopsies, treated tumors, low tumor fraction, necrosis, and limited tissue can therefore affect the component of HRD testing that depends on allelic imbalance and copy-number architecture—even when SNVs and small indels remain technically detectable.
“We can do a lot with a little, but we can’t do everything with a little” is more than a memorable quote. It captures an assay-design issue: a panel can have excellent small-variant sensitivity while remaining vulnerable in LOH, large deletion, copy-number loss, and genomic-scar reconstruction.
For laboratories, HRD is therefore a specimen-management service as much as it is an NGS assay. Selection of a larger untreated block, documentation of treatment history, pathologist estimation of tumor content, macrodissection, and an appropriate “quantity not sufficient” or qualified-result policy are central parts of performance.
5. Assays are not interchangeable
The Myriad strategy combines BRCA status with LOH, TAI, and LST. Foundation has historically used an LOH-centered approach. Other assays use different combinations, algorithms, training sets, and proprietary scales. “HRD-positive” on one system is not automatically the same measurement as “HRD-positive” on another.
The 2024 CAP review makes this especially relevant to Thermo Fisher. It describes:
Myriad GIS, with a clinically used cutoff of 42.
Foundation LOH, with a cutoff of 16 in the cited setting.
Oncomine Comprehensive Assay Plus’s genomic instability metric, a different 0–100 construct with a reported cutoff of 16.
Illumina TSO 500 HRD, which uses licensed Myriad GIS methodology.
A low-pass WGS/deep-learning approach from Sophia Genetics.
The fact that two metrics both use “16” does not make them comparable. Nor does high analytical concordance with an established assay automatically confer the same clinical validity for a particular drug and indication.
The 2022 paper recommends that publications and reports identify the assay, features measured, continuous score, cutoff, tumor type, and intended-use context. Titus demonstrates why that level of transparency is still necessary.
6. A genomic scar records history, not necessarily present function
This is perhaps the most important conceptual limitation.
Once LOH, TAI, LST, deletions, and chromosomal rearrangements have accumulated, the daughter cells inherit them. But a tumor may subsequently regain homologous-recombination function through a BRCA reversion mutation or reversal of BRCA methylation. The tumor can then become resistant to PARP inhibition while retaining the old genomic scars that produced its HRD-positive score.
A scar assay therefore asks:
Has this tumor lineage experienced HR deficiency?
The therapeutic question is closer to:
Is homologous recombination impaired in the currently dominant tumor population?
Those questions overlap, but they are not identical. This mismatch is particularly important after platinum or PARP exposure, when therapy has selected resistant clones. It also explains why a static, archival-tissue HRD result may become progressively less informative later in the disease course.
7. Epigenetic HRD is undermeasured
BRCA function may be lost through promoter methylation rather than sequence alteration. Many current assays do not directly measure methylation. Moreover, methylation can potentially be reversed, restoring function and producing early recurrence or platinum/PARP resistance despite the genomic-scar result.
Titus treats this as a major knowledge gap: the assay may detect the historical consequences without capturing the epigenetic mechanism or its subsequent reversal. For a sequencing company, this is a reminder that a DNA panel—even a very broad one—is not necessarily a complete HRD assay.
8. “HR-proficient” is becoming its own heterogeneous research category
Historically, most attention went to HRD because it supplied the therapeutic opportunity. Titus suggests that HRP tumors may now become equally important. HRP should not be treated simply as “nothing detected.”
Examples in the article include:
Strong CCNE1 amplification, associated with replication-fork stress and presented as evidence pointing away from the classic deletion/LOH-rich HRD phenotype.
RB1 loss in some HRP tumors, which may identify tumors behaving more like HRD tumors.
HRP tumors retaining RB1, which emerging data suggest may have particularly poor survival.
Possible differences in immune infiltration and PD-1/PD-L1 biology between BRCA-altered/HRD and HRP tumors.
These observations remain investigational. A CCNE1 amplification or RB1 result should not be promoted as a validated substitute HRD classifier. But they illustrate why comprehensive profiling may ultimately be more useful than an isolated positive/negative scar report.
9. Evidence is strongest in high-grade serous ovarian cancer
The field tends to talk as though HRD were a tumor-agnostic property. Yet both genomic-scar patterns and clinically appropriate thresholds may depend on tissue and histology. The strongest evidence remains in high-grade serous ovarian carcinoma, with some trial inclusion of endometrioid and other nonmucinous epithelial cancers.
Titus points to clear cell carcinoma and carcinosarcoma as areas where interesting cases exist but robust evidence does not. The 2022 paper makes the broader principle explicit: assays and cutoffs should be validated in their intended-use populations because the genomic manifestation of HRD may differ by tumor type.
What has changed from 2022 to 2026?
The 2022 paper described a harmonization problem: inconsistent definitions, different assay components, different thresholds, and inadequate reporting. The 2024 CAP review translated that framework into available laboratory technologies, including Oncomine Comprehensive Assay Plus.
Titus in 2026 shows three important developments:
The field now has more routine clinical experience, making tissue limitations, borderline scores, and intratumoral heterogeneity impossible to ignore.
Comprehensive profiling is beginning to place the HRD score in a larger biological context—BRCA and non-BRCA causes, CCNE1 amplification, RB1 status, resistance, and immune biology.
The static nature of genomic scars is increasingly recognized as a fundamental limitation, particularly after treatment and biological reversion.
In other words, the problem has evolved from “How do we standardize competing HRD assays?” to “Can a single HRD category adequately represent a heterogeneous and changing tumor?”
Some Technologies
Roswell Park uses an Oncomine Precision Assay for additional targeted tumor profiling while also relying on Myriad MyChoice CDx and FoundationOne CDx. That is an important distinction: the Oncomine panel can contribute information about HRR genes and wider tumor biology, but its presence in the workflow does not by itself make it interchangeable with the HRD companion diagnostic used in a pivotal trial.
Meanwhile, the article gives a competitor example: Labcorp’s OmniSeq Insight incorporates an optimized Illumina TSO 500 HRD workflow and combines genomic-instability measurement with broad DNA/RNA profiling and PD-L1. The market signal is that HRD is moving toward integrated comprehensive profiling, rather than surviving indefinitely as a narrow standalone score.
The strategic lessons are:
Do not market HRD as merely another panel output. The platform must distinguish causal HRR variants, genomic-scar measurement, and current functional status.
Treat the algorithm and clinical evidence as part of the product. Sequencing performance alone does not establish interchangeability with a trial-validated CDx.
Make preanalytics visible. Tumor fraction, treatment status, tissue area, copy-number confidence, and specimen-selection guidance are critical.
Retain and report the continuous metric. A binary call should not erase the actual score or its proximity to the cutoff.
Build a contextual report. BRCA findings, relevant non-BRCA genes, possible biallelic status, genomic-instability components, strong counter-signals such as CCNE1 amplification, and applicable limitations should be separable.
Validate by tumor type and intended treatment use. An ovarian-cancer threshold cannot simply be carried into prostate, pancreatic, breast, or other tumors because the same biological pathway may leave different genomic patterns.
Plan for longitudinal disease. Static scar testing is strongest near diagnosis in untreated tissue. Recurrent, post-PARP, or platinum-resistant disease may require rebiopsy, liquid-biopsy resistance testing, methylation assessment, or eventually functional HR measurement.
Avoid overclaiming the HRP category. A negative scar score is not proof of homogeneous normal repair biology, and HRP itself may contain clinically distinct subgroups.
A concise teaching conclusion
The cleanest conclusion for the executives is:
HRD testing attempts to infer a dynamic DNA-repair phenotype from a mixture of causal variants and historical genomic damage. The available assays differ in what they measure, how they score it, where they set the cutoff, and the clinical trials to which that cutoff is tied. Tissue quality, tumor heterogeneity, non-BRCA biology, methylation, and acquired BRCA reversions can all separate the reported HRD category from the tumor’s current therapeutic vulnerability. The next generation of testing will therefore need to move beyond a solitary positive/negative scar score toward integrated, tumor-specific, and eventually longitudinal assessment.
That, I think, is the real 2026 CAP Today message: HRD is a valuable biomarker, but “HRD-positive” is still a compressed summary of a much more complicated biological and measurement problem.
