Search Results

The clonoSEQ Assay is an in vitro diagnostic that uses multiplex polymerase chain reaction (PCR) and next-generation sequencing (NGS) to identify and quantify rearranged IgH (VDJ), IgH (DJ), IgK and IgL receptor gene sequences, as well as translocated BCL1/1gH (J) and BCL2/1gH (J) sequences in DNA extracted from bone marrow from patients with B-cell acute lymphoblastic leukemia (ALL) or multiple myeloma (MM), and blood or bone marrow from patients with chronic lymphocytic leukemia (CLL).

The clonoSEQ Assay measures minimal residual disease (MRD) to monitor changes in burden of disease during and after treatment. The test is indicated for use by qualified healthcare professionals in accordance with professional guidelines for clinical decision-making and in conjunction with other clinicopathological features.

The clonoSEQ Assay is a single-site assay performed at Adaptive Biotechnologies Corporation.

The clonoSEQ Assay is a next-generation sequencing (NGS) based assay that identifies rearranged IgH (VDJ), IgH (DJ), IgK, and IgL receptor gene sequences, as well as translocated BCL1/IgH (J) and BCL2/IgH (J) sequences. The assay also includes primers that amplify specific genomic regions present as diploid copies in normal genomic DNA (gDNA) to allow determination of total nucleated cell content.

Testing begins with gDNA extracted from the specimen supplied (Figure 1). Extracted gDNA quality is assessed and rearranged immune receptors are amplified using a multiplex PCR. Reaction-specific index barcode sequences for sample identification are added to the amplified receptor sequences by PCR. Sequencing libraries are prepared from barcoded amplified DNA, which are then sequenced by synthesis using NGS. Raw sequence data are uploaded from the sequencing instrument to the Adaptive analysis pipeline. These sequence data are analyzed in a multi-step process: first, a sample's sequence data are identified using the sample index sequences. Next, data are processed using a proprietary algorithm with in-line controls to remove amplification bias. When the clonoSEQ Clonality (ID) assessment is conducted, the immune repertoire of the sample is checked for the presence of DNA sequences specific to "dominant" clone(s) consistent with the presence of a lymphoid malignancy. Each sequence that is being considered for MRD tracking is compared against a B cell repertoire database and assigned a uniqueness value that, together with its abundance relative to other sequences, is used to assign the sequence to a sensitivity bin which will be used in the estimation of the reported LoD and LoO on the patient report. During clonoSEQ Tracking (MRD) assessment, the complete immunoglobulin receptor repertoire is again assessed, and the previously identified dominant clonotype sequence(s) are detected and quantified to determine the sample MRD level. The clonoSEQ Assay MRD assessment measures residual disease in a biologic sample.

Here's a breakdown of the acceptance criteria and study details for the Adaptive Biotechnologies clonoSEQ Assay, based on the provided FDA 510(k) summary:

This device, the Adaptive Biotechnologies clonoSEQ Assay, is an in vitro diagnostic (IVD) that identifies and quantifies rearranged immune receptor gene sequences (IgH, IgK, IgL) and translocated BCL1/IgH and BCL2/IgH sequences using multiplex PCR and Next-Generation Sequencing (NGS). It measures Minimal Residual Disease (MRD) in patients with B-cell acute lymphoblastic leukemia (ALL), multiple myeloma (MM), and chronic lymphocytic leukemia (CLL) to monitor disease burden. The current submission is an expansion of indications to include blood samples from CLL patients.

1. A table of acceptance criteria and the reported device performance

The provided document doesn't explicitly list "acceptance criteria" in a single table, but rather describes the performance characteristics that were measured and the outcomes for each. I will compile these for the CLL in Blood indication, as this is the focus of the 510(k) expansion.

Acceptance Criteria & Reported Device Performance for clonoSEQ Assay (CLL in Blood)

2. Sample sizes used for the test set and the data provenance

• Precision Studies (Test Set):
  • CLL in BMA: 22 patients. Contrived samples (blending patient gDNA with healthy donor BMA gDNA). 360 contrived samples tested, yielding ~7,480 MRD measurements.
  • CLL in Blood: 15 patients. Contrived samples (blending patient gDNA with healthy donor blood gDNA). 320 contrived samples tested, yielding ~4,785 MRD measurements.
• Linearity Studies (Test Set):
  • Cell Lines in BMA: 3 CLL cell lines (HG-3, MEC-1, PGA-1). Blended cell line gDNA with healthy subject gDNA. Data shown for multiple DNA inputs and 11 MRD frequencies.
  • Clinical BMA Specimens: Re-analysis of data from the 22 CLL patients in the precision study.
  • Clinical Blood Specimens: Re-analysis of data from the 15 CLL patients in the precision study.
• Accuracy (Concordance with mpFC in Blood Clinical Samples): 299 matched clinical samples.
• Limit of Blank (LoB): 22 CLL patient samples (for trackable sequences) and healthy bone marrow samples. For blood, 15 CLL patient samples and healthy blood samples.
• Limit of Detection/Quantitation (LoD/LoQ): 22 CLL patient specimens (BMA) and 15 CLL patient specimens (Blood). Contrived dilution series.
• Interfering Substances: 4 different donors for both BMA and blood samples. Each condition replicated 8 times. Additional assessment on 4 CLL clinical blood specimens.
• Cross-Contamination/Carryover:
  • Automated DNA Extraction (BMA/BMMC): Panel of 6 lymphoid malignancy cell lines (3 ALL, 3 MM), 10% spiked into healthy BMA/BMMC pool + PBS blanks.
  • Automated DNA Extraction (Blood): Panel of 6 lymphoid malignancy cell lines, 10% spiked into normal healthy blood + PBS blanks.
  • PCR, Library Pooling, Sequencing: gDNA from blood from healthy subjects (MRD-negative) and 5% spiked cell line gDNA blends.
• Clinical Studies (to support prognostic utility):
  • NCT02242942 (Primary CLL study): Samples and outcomes data from 445 patients initially. For primary analysis, 337 patients had usable clonoSEQ Assay MRD data and clinical outcomes (after QC and excluding early progression).
  • NCT00759798 (Secondary CLL study): 111 front-line CLL patients with clonoSEQ ID samples, 137 clonoSEQ MRD samples (also evaluated by 4-color flow cytometry). Bone marrow available for 75 patients, blood for 62 patients (26 had both). 3 patients excluded due to missing clinical covariates.

Data Provenance:
The data provenance is largely implied as originating from Adaptive Biotechnologies Corporation as a single-site assay performed in Seattle, Washington. The studies involve clinical specimens from patients with CLL.

• The clinical validation studies reference specific US clinical trials (NCT02242942, NCT00759798), indicating prospective data collection from multi-center clinical trials, likely including locations within the US.
• The analytical studies (precision, linearity, etc.) used both contrived samples (blending patient/cell line gDNA with healthy donor gDNA) and clinical specimens, processed and analyzed at the Adaptive Biotechnologies lab. These are described as analytic validation studies.

3. Number of experts used to establish the ground truth for the test set and the qualifications of those experts

The document describes an in vitro diagnostic (IVD) test that quantifies specific DNA sequences. The "ground truth" for the analytical performance studies (precision, linearity, LoD/LoQ) is established by the known concentrations of the contrived samples, where specific amounts of malignant cells or gDNA from patient samples (with known clonal sequences) are blended into a background of healthy donor gDNA. This is a common and appropriate method for analytical validation of quantitative IVD assays.

For the Clinical Studies, the "ground truth" for patient outcomes is clinical progression-free survival (PFS) data and disease assessment from the clinical trials (NCT02242942 and NCT00759798). These outcomes are established by the clinical trial investigators, who would be qualified healthcare professionals (e.g., oncologists, hematologists) following standard clinical practice and trial protocols. The document does not specify a number of "experts" to establish ground truth in the sense of independent expert review of images or clinical cases, as would be common for an imaging AI device. The ground truth for this device is based on quantifiable molecular levels and patient outcomes.

4. Adjudication method for the test set

Not applicable in the typical sense for this type of IVD device. "Adjudication" usually refers to a process for resolving discrepancies in expert interpretations (e.g., radiologist reads).

• For the analytical studies, the ground truth is established by the known input concentrations of contrived samples or the re-analysis of patient data to show linearity. Reproducibility and accuracy are assessed by repeatedly measuring these known inputs across different conditions (operators, instruments, reagents, etc.).
• For the clinical studies, the ground truth is patient outcome data (PFS) collected as part of the clinical trials, which would be managed and reviewed according to standard clinical trial protocols, not a specific "adjudication method" as seen in image reading studies.

5. If a multi reader multi case (MRMC) comparative effectiveness study was done, If so, what was the effect size of how much human readers improve with AI vs without AI assistance

No, an MRMC comparative effectiveness study was not done. This is an in vitro diagnostic (IVD) assay, not an AI imaging device where human readers interact with AI. The clonoSEQ Assay is a laboratory test that provides quantitative MRD measurements. There is no "human reader" in the sense of interpreting an AI output directly influencing their decision-making for a specific case (like interpreting a radiology image with or without AI assistance). The output is a numerical MRD value reported to qualified healthcare professionals for use in clinical decision-making.

6. If a standalone (i.e. algorithm only without human-in-the-loop performance) was done

Yes, the "standalone" performance is the core of the studies presented. The clonoSEQ Assay is an algorithm-driven test that provides a quantitative MRD value. The precision, linearity, LoD/LoQ, and analytical specificity studies directly evaluate the "algorithm only" performance (i.e., the performance of the assay system independent of clinical interpretation for a specific patient). The output (MRD value) is generated by the assay system and its bioinformatics pipeline, which includes proprietary algorithms. The clinical studies then demonstrate the prognostic utility of this standalone quantitative output.

7. The type of ground truth used (expert consensus, pathology, outcomes data, etc)

The ground truth for the different aspects of the study include:

• Analytical Studies (Precision, Linearity, LoD/LoQ, Interfering Substances, Cross-Contamination): Known input concentrations of contrived samples (blending gDNA from patient/cell lines with healthy donor gDNA). This is a quantitative ground truth.
• Accuracy (Concordance with mpFC): Flow Cytometry (mpFC) results served as a comparator method. While not a "gold standard" pathology ground truth, it's a widely accepted method for MRD detection, and the study highlights clonoSEQ's greater sensitivity.
• Clinical Studies: Patient Outcomes Data, specifically Progression-Free Survival (PFS), as recorded during controlled clinical trials.

8. The sample size for the training set

The document does not explicitly state a "training set" size for the clonoSEQ assay's algorithmic development. For IVD assays based on NGS and bioinformatics pipelines, the "training" (or development/optimization) often involves extensive analytical validation, optimization using synthetic controls, and iterations on algorithms rather than a distinct "training set" like in machine learning for image classification.

The document states:

• "Raw sequence data are uploaded from the sequencing instrument to the Adaptive analysis pipeline. These sequence data are analyzed in a multi-step process: first, a sample's sequence data are identified using the sample index sequences. Next, data are processed using a proprietary algorithm with in-line controls to remove amplification bias."
• "When the clonoSEQ Clonality (ID) assessment is conducted, the immune repertoire of the sample is checked for the presence of DNA sequences specific to "dominant" clone(s) consistent with the presence of a lymphoid malignancy. Each sequence that is being considered for MRD tracking is compared against a B cell repertoire database and assigned a uniqueness value..."

This suggests the algorithms were likely developed and refined using a combination of synthetic data, known biological samples, and potentially retrospective patient data to build the "B cell repertoire database" and optimize the bias correction and clonality assessment. However, the exact sample sizes for this development phase are not provided in this 510(k) summary, as it primarily focuses on the validation of the finalized assay for regulatory approval.

9. How the ground truth for the training set was established

As inferred above, if there was a "training set" in the context of algorithm development, the ground truth would have been established through:

• Known molecular constructs/synthetic controls: For optimizing sequencing and amplification bias correction.
• Well-characterized cell lines and patient samples: Where the presence and frequency of specific clonal rearrangements are either known or confirmed by orthogonal methods (e.g., flow cytometry, Sanger sequencing, or other molecular techniques) to build the B cell repertoire database or tune the clonality determination.
• Expert knowledge of immunology and genetics: To design the algorithms that identify and quantify rearranged gene sequences and interpret their significance in the context of hematological malignancies.

The document implicitly refers to this through descriptions of the "proprietary algorithm with in-line controls" and comparison against a "B cell repertoire database," indicating an internally developed and optimized system for which ground truth would have been internally established during its development phase.

Ask a specific question about this device