With the recent shift in the industry as the result of this decision, a common question is “What lies in store for digital pathology system manufacturers in the wake of the FDA granting Philips’ de novo classification request for its IntelliSite device?”
First, What is De Novo?
De novo classification is the regulatory pathway for marketing clearance for novel, low-to-moderate-risk medical devices that are the first of their kind. Devices that are classified through the de novo process may be marketed and used as predicates for future 510(k) submissions. Through the Philips de novo, the FDA classified that whole slide imaging devices intended for primary diagnosis as Class II medical devices subject to Premarket Notification (510(k)) and other special controls. This classification will apply to all subsequently cleared WSI devices that use IntelliSite as a predicate.
Through the product code (PSY) that the FDA created as part of the de novo classification, the FDA defines a WSI system as consisting of an image management system, a scanner with associated software, and a display monitor. It explicitly excludes automated image analysis. Presumably, this exclusion applies to primary diagnosis only; it should not preclude automated image analysis of IHC stained slides for secondary (adjunctive) diagnosis, which the FDA has cleared in the past using other product codes (NQN, NOT). In other words, there is no known regulatory barrier to a manufacturer seeking a clearance that couples manual digital reads for primary diagnosis with image analysis for secondary diagnosis. The language of the Indications for Use for this scenario may require some negotiating and fine-tuning with the FDA, but it is not an insurmountable impediment.
The FDA granted a general indication to the Philips device, and thus to the device type. The indications do not specify tissue type or stain/stainer, but they do specify tissue specimen, viz. FFPE. Frozen section, cytology and non-FFPE specimens are formally contraindicated. The removal of a contraindication (or the inclusion of what was previously contraindicated in the Indications for Use) always triggers a new 510(k) for a Class II device, and could result in another de novo submission if the FDA concludes that it renders the candidate device not substantially equivalent to its predicate. I do not know if the contraindications resulted from Philips merely excluding these specimen types from their studies (e.g. due to time constraints), or if the FDA has concerns regarding the safety & effectiveness of using WSI systems to render diagnoses from them.
The de novo classification order also established the special controls that apply to WSI for primary diagnosis. These special controls comprise a published technical guidance for device design and verification (including performance testing), specific labeling requirements, and requirements for design validation studies. The studies include a usability (human factors) study, and analytical studies for both accuracy (concordance) and precision.
Manufacturers seeking 510(k) clearance for their WSI systems intended for primary diagnosis must comply with and fulfill the requirements of the published special controls, and mimic the design of the analytical studies of the predicate device (IntelliSite) as closely as possible. The bar for the analytical studies is especially high, as Philips utilized 4 sites, 27 pathologists, 2000 cases, 3400 slides (covering 20 organs and 54 sub-types), and 16,000 reads.
Do not expect the FDA to lower this bar for the foreseeable future. In time, the FDA may grow more comfortable with the performance and safety & effectiveness of the technology, from both pre-market (e.g. multiple successful 510(k) clearances) and post-market (e.g. post-market surveillance data from adverse events and recalls) perspectives. I think it will take at least two years for the FDA to reach any level of comfort with WSI systems, and probably longer.
So, What Happens Now?
Manufacturers will continue to push the envelope now that WSI devices are cleared for primary diagnosis in the USA. What innovations could trigger another de novo, or even a PMA? Some possibilities are:
- An image analysis algorithm intended for primary diagnosis (e.g. for H&E stains)
- Whole tissue-based image analysis algorithms, as opposed to the ROI/FOV-based algorithms that currently have clearance
- Any functionality that would significantly reduce the role of the pathologist in interpreting images or rendering a diagnosis. The special controls make it very clear that the FDA expects the pathologist to play an active and critical role in the use of a WSI intended for primary diagnosis, mandating requirements for including such disclaimers in the Indications for Use and other labeling.
- Any expansion of indications to include what is contraindicated in the predicate device, viz. slides prepared from frozen tissue, cytology, and non-FFPE hematopathology specimens
- The development of a phantom or other medical device development tool for the purpose of eliminating or significantly altering the design of one or more of the required study types. In this context, a phantom would be some mechanism by which image quality can be objectively assessed (resolution, clarity, focus, colorimetric fidelity, etc.).
Speaking of Medical Device Development Tools…
The FDA defines a Medical Device Development Tool (MDDT) as a scientifically validated tool that aids device development and regulatory evaluation. The FDA MDDT qualification program is currently in pilot status, but the objective of the process is to qualify tools for a specific context of use, which medical device manufacturers can use to support their premarket submissions without having to independently validate them.
A WSI system manufacturer utilizing a previously qualified MDDT (as opposed to validating it as part of a premarket submission) could mean the difference between a 510(k) clearance and the need for another de novo classification submission.
If the FDA accepts a qualified or validated phantom, which analytical studies could it eliminate or significantly alter? I do not think that a phantom replaces the need to compare the outputs of image analysis algorithms to themselves (algorithm precision). As long as the reads from the optical microscope are the ground truth, then I do not see the phantom obviating the need to establish the concordance of diagnoses (accuracy), either for manual reads or image analysis. The phantom may, however, replace the comparison of optical and digital manual identification of histopathology features as the means to validate imaging precision for manual reads, making such precision evaluation more objective.
Note that there is an FDA qualification in progress for a MDDT in the digital pathology domain. The tool aims to reduce variability in imaging precision studies by enabling different pathologists to evaluate the same FOVs in both digital and microscope modes. Presumably, this tool could be used for evaluating the precision of manual reads (digital to optical) and image analysis (algorithm outputs to manual reads). Link: https://nciphub.org/groups/wsi_working_group Read more about OptraSCAN On-Demand Digital Pathology: www.optrascan.com. Disclaimer: In seeking to foster discourse on a wide array of ideas, the Digital Pathology Association believes that it is important to share a range of prominent industry viewpoints. This article does not necessarily express the viewpoints of the DPA, however we view this as a valuable point with which to facilitate discussion.