(90 days)
QbTest is indicated to be used to aid in the clinical assessment of ADHD. QbTest results should be interpreted by qualified health care professionals only.
QbTest provides clinicians with objective measurements of hyperactivity, impulsivity, and inattention to aid in the clinical assessment of ADHD. QbTest results should be interpreted only by qualified professionals.
QbTest is a non-invasive test that has been developed to provide precise quantitative assessment of the capacity for an individual to pay attention to visual stimuli and inhibit impulses. There are three cardinal disturbances in Attention-Deficit Hyperactivity Disorder (ADHD); impaired attention, hyperactivity and impulsivity. QbTest provides an accurate and reproducible measure of an individual's capacity in each of these three domains by utilizing a consistent challenge paradigm coupled with detailed real-time measurements of behavior and performance. The fundamental core of QbTest is a computer-assisted attention and impulse control task and simultaneous recording of activity using an infrared camera for motion measurements.
The system consists of the following components:
- Client software
- Responder button (also referred to as responder unit)
- Infrared camera
- Reflective motion marker
- User manual
- Technical manual
- Stimulus card
- Camera stand
- Measuring tape
- QbTest Behaviour Rating Scale
- In addition, the user must have access to a remote server that generates test reports
The provided FDA 510(k) summary for QbTest v3.5 is a predicate device comparison, rather than a typical AI/ML medical device submission with specific performance acceptance criteria for an algorithm. Therefore, the information typically found for AI/ML device validation studies (like sensitivity/specificity, ROC curves, MRMC studies, precise ground truth establishment for a test set, etc.) is largely absent.
The submission focuses on demonstrating substantial equivalence to a previously cleared device (QbTest K040894) and the Gordon Diagnostic System (K854903) by showing similar intended use, technological characteristics, and safety/performance based on normative data collection and prior published clinical studies of the device and its predecessor.
Here's an attempt to answer your questions based on the provided text, highlighting what is present and what is not:
1. Table of Acceptance Criteria and Reported Device Performance
Strict "acceptance criteria" as you'd find in an AI/ML device validation (e.g., minimum sensitivity or specificity) are not stated in this 510(k) summary. The performance is demonstrated through:
| Acceptance Criteria (Implied) | Reported Device Performance |
|---|---|
| Demonstrated safety and effectiveness (as per predicate) | System tested to EN60601-1 and EN60601-1-2 standards. |
| Provision of objective measurements for hyperactivity, impulsivity, and inattention | QbTest provides measurements in these domains. |
| Aid in clinical assessment of ADHD | Four published studies evaluated clinical validity. |
| Reliability (test-retest consistency) | Two test-retest studies completed. |
| Normative data for interpretation (age/gender-specific) | Normative database of 1307 individuals (6-60 years). |
2. Sample Size Used for the Test Set and Data Provenance
- Sample Size for Clinical Validation/Normative Data:
- "Normative tests have been gathered from several different cohorts resulting in a normative database of 1307 individuals between 6 and 60 years with an even age and gender distribution." This 1307-individual dataset serves as the primary "reference" or "test set" against which individual patient performance is compared. It's not a "test set" in the sense of an independent validation set for algorithm performance, but rather a normative reference.
- The submission also mentions "four published studies which have evaluated the clinical validity of the QbTest for its intended use population" and "two test-retest studies." The individual sample sizes for these specific studies are not provided in this summary.
- Data Provenance: Not explicitly stated (e.g., country of origin, retrospective/prospective). However, the general nature of normative data collection often implies a prospective or at least a systematic retrospective collection from a defined population. The submitter is Swedish (Qbtech AB, Stockholm, Sweden), which might suggest some data from that region, but this is speculative.
3. Number of Experts Used to Establish Ground Truth and Their Qualifications
- Ground Truth for QbTest: The QbTest itself generates the objective measurements of hyperactivity, impulsivity, and inattention. The "ground truth" for ADHD diagnosis isn't established by individual experts reviewing test data; rather, the test aids qualified healthcare professionals in making the diagnosis.
- The "normative database" would have been collected from individuals (both with and without diagnosed ADHD, presumably) where their diagnostic status would have been established by qualified clinicians, but the number and qualifications of these clinicians are not specified.
- The "four published studies" and "two test-retest studies" would have involved clinical professionals to manage and interpret the data, but no specific count or qualifications are provided in this summary.
4. Adjudication Method for the Test Set
- Not applicable in the typical sense of expert review for an algorithm's output. The QbTest itself produces quantitative output. Any diagnostic "ground truth" used in the underlying clinical studies (if those studies involve comparing QbTest outputs to clinical diagnoses) would likely follow standard clinical diagnostic procedures, which may involve adjudication, but this is not described in the 510(k) summary.
5. Multi Reader Multi Case (MRMC) Comparative Effectiveness Study
- No, an MRMC study is not mentioned or described. This type of study (human readers with and without AI assistance) is typically performed for imaging or diagnostic algorithms that directly influence a human reader's interpretation. QbTest provides quantitative data that aids a clinician, rather than directly modifying their interpretation of, for example, an anatomical image.
6. Standalone (Algorithm Only Without Human-in-the-Loop Performance) Study
- Yes, in essence. The QbTest itself is a standalone device that measures parameters (hyperactivity, impulsivity, inattention). Its performance is assessed by how well these measurements are collected and how consistently they reflect a person's behavior/performance. The "clinical validity" studies assess the utility of these measurements in aiding ADHD assessment, which is analogous to a standalone performance evaluation of the device's output. The summary refers to "four published studies which have evaluated the clinical validity of the QbTest for its intended use population."
7. Type of Ground Truth Used
- Clinical Diagnosis/Phenotype (implicit): For the "clinical validity" studies, the "ground truth" would likely be a clinical diagnosis of ADHD (or lack thereof) made by qualified professionals, following established diagnostic criteria (e.g., DSM criteria). The QbTest's output is then correlated with or assessed for its ability to discriminate based on this clinical "ground truth."
- Observed Behavior/Performance (inherent): For the test-retest reliability studies, the ground truth is the inherent stability of an individual's performance and behavior on the test over time.
- Normative Data: The "normative database" itself serves as a "ground truth" for what is considered typical performance for a given age and gender.
8. Sample Size for the Training Set
- Not Applicable / Not Explicitly Stated. The QbTest described is a direct measurement system, not a machine learning algorithm that is "trained" on a dataset in the conventional sense. The "normative database" of 1307 individuals functions more like a reference set rather than a "training set" for an AI model.
9. How the Ground Truth for the Training Set Was Established
- Not Applicable / Not Explicitly Stated. As it's not an ML training set, the concept of establishing ground truth for training doesn't apply directly. The "normative database" was established by collecting data from "several different cohorts" of individuals between 6 and 60 years old. The methods for this collection are described as being in the "technical manual," but not detailed here. Presumably, these were "healthy" or "typically developing" individuals to establish the "norm."
Summary of Limitations based on the provided text:
This 510(k) summary is typical for a non-AI/ML device that is seeking clearance based on substantial equivalence to an existing predicate. It heavily relies on prior clearance and existing clinical literature demonstrating the utility of the type of device. It does not provide the detailed information about AI/ML validation studies that are now common in submissions for software as a medical device (SaMD) utilizing AI/ML algorithms.
N/A