Ever since Harvard professor Clayton M. Christensen came up with the concept of disruptive innovation in 1997, technological disruption has occurred across all industries. Life sciences are no exception, and digital transformation continues to impact almost every facet of healthcare. For example, Software as a Medical Device (SaMD) use cases now run the gamut from patient-provider interactions to image analysis, personalized treatments, pharmaceutical discoveries, and multiple other functions. Here are some of the challenges facing manufacturers in their quest for regulatory approval of SaMDs.

The Role of AI and ML in SaMD

The increasing role of artificial intelligence (AI) and machine learning (ML) in SaMDs presents challenges for the European Union’s medical device regulations (EU MDR). Uses of these technologies range from simple algorithms to complex, cognitive computing, big data analytics, deep learning, speech and image recognition, natural language processing, and robotics process automation, among others. 

Some smart technologies featured in SaMD products include:

• AI-driven digital image processing, restoration, and analysis of CT scans. Examples of this technology include orthopedic planning software, skin disease detection programs, and diabetic retinopathy detection. 
• Identification and monitoring of disease patterns and progression. Examples include ECG signal analysis, AI-enhanced wearables, and patient monitoring systems, such as blood glucose and hypertension monitors. These require Internet of Things (IoT) integration to deliver insights gleaned from real-time data, raise alarms before a disease develops or progresses, and improve disease management. 
• Oversight of prescription trends and their correlation with patients’ brand-seeking medication behaviors, and the deployment of smart robotic devices to dispense medication, sterilize equipment, and track medication adherence.  

AI and ML-based SaMDs have the potential to adapt and optimize device performance in real time as data becomes available for incorporation. The technology has recently been integrated into robotic surgical devices to deliver advanced decision-making and precision in performing delicate procedures.

EU MDR’s SaMD Approval Approach

Of course, regulatory approvals are necessary to use this technology in a real-world patient setting. The EU MDR doesn’t take the same approach to SaMDs as the FDA, where the use of a predicate device with substantial equivalence can get approval fast-tracked. The EU MDR has a separate framework with 22 rules to follow (EU Medical Device Regulation (MDR 2017/745) to obtain a classification output, get a medical device approved, and acquire a Conformitè Europëenne (CE) mark. 

In Europe, devices with integrated software are referred to as medical device software (MDSW). These products fall into one of four risk classifications, which are low risk (Class 1a), medium risk (Class IIa), medium-high risk (Class IIb), or high risk (Class IIIa). 

An MDSW can be released to market in one of two ways:

1. As a medical device in its own right, or 
2. As a component of any other medical device. 

The first option requires manufacturers to follow a specific, in-depth regulatory process that includes a conformity assessment to determine whether the medical device meets EU MDR requirements. However, the second option is much less rigorous. It can lead to an MDSW being released to market as a component of another device under a conformity assessment that applies to the parent device. 

EU-Specific Challenges Facing SaMD Approvals

One issue impacting the EU MDR’s SaMD approval pathways is that the current framework doesn’t specifically address AI’s self-learning abilities. ML algorithms are generally regarded as medium risk (Class IIa), which doesn’t take into account the rapid, iterative nature of AI systems capable of decision-making. 

Although manufacturers can “self-declare” Class I devices, it doesn’t mean SaMDs can get to market faster in Europe. Under the EU MDR, many software devices have been up-classified, which means they now require the involvement of a Notified Body. The result of this scenario is often a 3-6 month delay in the review and approval process. 

Then, of course, the EU’s AI Act (AIA) will likely influence all AI-based SaMDs that interact with humans and have a potentially far-reaching international impact. In addition, the EU MDR is compounding the challenges facing SaMD providers by demanding they resubmit all devices already in the market for compliance purposes. Grace periods for resubmission depend on a device’s risk classification, with Class I devices having until May 2024 to comply. 

AI and ML-Based SaMD Benefits for Providers and Patients

Despite the challenges presented by the implementation of EU MDR’s approval processes, AI and ML-based SaMDs offer significant benefits that make them worth the effort. 

• Integrating AI and ML enables the automation of everyday tasks and optimizes the time providers spend with patients. 
• Rapid analysis of big datasets delivers insights faster and at a lower cost. 
• Many SaMDs integrate with electronic medical records, reducing the administrative workload and minimizing the risk of human error. 

Since 30% of healthcare costs apply to administrative processes, reducing labor hours also means substantial savings in time and costs. 

The Final Analysis

Using AI and ML-based SaMDs enables healthcare practitioners to make better clinical decisions, provide better care, and boost patient retention. It can improve the morale and productivity of medical practitioners and staff by enabling them to spend more time caring for patients and improve treatment outcomes. Although several challenges exist regarding cybersecurity, privacy concerns, and the safety of AI-led decision-making, compliance with EU MDR approval processes offers a high level of reassurance.   

LSI Europe ‘22 Emerging Medtech Summit will be hosting a panel session to cover the changes of EU MDR in greater detail. Be sure to register at: https://lsi-medtech.com/LSIEurope22