What are the three types of clinical evaluation data you need to certify a software medical device?.

Understanding clinical data is the basis of a comprehensive clinical evaluation process to support software medical device certification

To achieve certification and take a software medical device to market, manufacturers need to implement a variety of regulatory processes. Key among them is clinical evaluation. A robust clinical evaluation process allows manufacturers to demonstrate that their software device is safe and performant when used as intended.

For that you need clinical data. Manufacturers can identify existing clinical data to use, or generate new clinical data to support their device, and during the device’s lifetime manufacturers should build and maintain a body of evidence with high-quality clinical data to support its ongoing use.

A simple framework

At Scarlet, we find that clinical evaluation is often the regulatory process that causes manufacturers the most concern. One of the key reasons is that it can be difficult to figure out what kind of clinical data is required to support a device.

Fortunately, for software medical devices, there is a simple framework that can guide manufacturers towards a thorough and coherent clinical evaluation process.

When it comes to software medical devices, there are three different types of clinical data that a manufacturer needs to provide:

1. Technical performance data
2. Clinical association data
3. Clinical performance data

To get a device certified, a manufacturer will need to provide data in all three categories. So what exactly do they mean?

The three types of clinical data

Let’s consider a device that analyses eye scans and categorises the blood vessels in the back of the eye (the retina) based on how crooked (or tortuous) they are. The device then provides a risk score that can be used to determine someone’s risk of developing cardiovascular disease.

The inputs to the device are an eye scan and some basic patient demographic information (like age and sex). The output is a score to indicate the likelihood of cardiovascular disease.

• 1. Technical performance

It’s important that the score provided by the device is consistent and reliable if it’s given the same type of inputs - this is the essence of technical performance.

The manufacturer can generate technical performance data by confirming that the output score is consistent for eye scans and patients with similar attributes.

• 2. Clinical association

The output of the device is predominantly based on the ‘crookedness’ of the blood vessels at the back of the eye. An interesting insight! But is this a well-established relationship? This is the basis of clinical association.

The manufacturer can conduct a search of the scientific literature and identify clinical association data from prior research that demonstrates a connection between ‘retinal vessel crookedness’ and ‘cardiovascular disease’. So far, so good.

• 3. Clinical performance

Finally, the manufacturer hopes that the risk scores provided by the device will be helpful in diagnosing cardiovascular disease. One of the device’s clinical benefit claims is that it can detect the presence of cardiovascular disease with a high degree of specificity. Whether the device, when used as intended, can actually achieve this clinical benefit is referred to as the clinical performance.

The manufacturer can generate clinical performance data by comparing the output of the device, when used as intended, with a gold-standard diagnostic measure, and analyse the rate of false positives and false negatives generated by the device.

The foundation of certification

While there are many other components to a high-quality clinical evaluation process, if a manufacturer is able to apply this framework and provide high-quality data covering the three areas of technical performance, clinical association, and clinical performance, they are well on their way to building a comprehensive clinical evaluation process that should support the certification of their software medical device.

You'll find more on the clinical data framework outlined above in this guidance document.