FDA Technical Guidance Document Does Not Go Far Enough

As many readers are aware, on April 20th the FDA released its Technical Guidance Document for Digital Pathology Whole Slide Imaging Devices after draft proposal/review from February of last year.

Given its title, this is a technical document, obviously. However, in my opinion, the document, beginning with draft to this finished product, I am sure despite comments to the contrary about important factors that need to be considered by manufacturers, does not go far enough addressing human factors.

While there are pages and pages about technical considerations dealing with slide input, scanning, magnification, luminescence, color reproducibility, image manipulation and computer environment and so on and so on that you can read in the document, there is a clear omission of the human element.

However, it seems to me since the first public meeting in Maryland in 2008 I believe and subsequent presentations by FDA representatives at public meetings, there have always been, as appropriate, consideration of the human factors involved.

Technical factors provide one can reliably, reproducibly scan slides as represented in their “native” form with all the pieces and folds and wrinkles and cracked cover slips and poorly stained H&E sections, etc… and get a photocopy of that slide someone can read, the scanner has done its job.

We have that now.  A 40x scan of the entire 3×1 glass slide and the devices create a diagnostic quality image. A lot of consideration of input but little consideration of output.  The human factor.

And here in lies the problem.  The technology is sound. The input qualifications specified.  The output characteristics less so and the human factor, a deviation from the past, not even considered in the context of the technical factors.

This is the bottom line.  A hard case is a hard case…Hard to do reproducibility studies with a different method when the “gold standard” has inherent issues. In this analogy, digital pathology should be held to same standard as “gold standard” microscope. So, if 12 people agree on the diagnosis together 75% of the time – the new(er) method should meet or exceed 75%.  Ditto if 95% or 50%.  This is what makes it a challenge.  Cases cannot be too “easy” or too “hard”.

If you show a case to 3 pathologists you may get 4 opinions.

Anyone heard about the surgeon who wanted a one-armed pathologist? He asked the hospital for it because he said the pathologist always says “On one hand it could be this, or on the other hand it could be this.”

What if a pathologist gets the same diagnosis wrong using the microscope and whole slide imaging — this may be regarded as concordant diagnoses.  Just as if I look at a case with rose-colored glasses and get the diagnosis wrong as I did without the glasses.  One could argue this is a pathologist/observer/reader problem – not the scan, monitor or glasses or color. And it is a human problem.  Not a scanner or image problem but we have controlled already for many of the technical issues. The professional issues, if we look at this like a lab test with a technical component (TC) and professional component  (PC) are more difficult to reproduce and measure beyond the instrumentation.

Here is the other bottom line.  I am not sure the FDA is really going to do anything with this until and unless there is a market where there are applications that provide benefit and therefore the potential for risk.  If someone were to start to do 100% digital reads or close to it as our colleagues in Europe have in recent years the FDA would have the impetus to regulate this.  It is a catch 22. A circular argument. But one that needs to look at not only technical issues but professional ones as well.