Letter to the Editor: Can Machine Learning Algorithms Predict Which Patients Will Achieve Minimally Clinically Important Differences From Total Joint Arthroplasty?

To the Editor,

Kudos to Fontana and colleagues [3] for their excellent paper, “Can Machine Learning Algorithms Predict Which Patients Will Achieve Minimally Clinically Important Differences From Total Joint Arthroplasty?”.

I am writing because I just took a look at some of the papers cited in the related Take 5 interview [5] and encountered what may be the second-most understated sentence in all of the medical literature.

Of course, the most understated sentence a medical paper is from the opening paragraph of “A Structure for Deoxyribose Nucleic Acid” by Watson and Crick in Nature [7]: “We wish to suggest a structure for the salt of deoxyribose nucleic acid. This structure has novel features which are of considerable biological interest” [7].

But a sentence in the abstract from, “Do Surgeon Expectations Predict Clinically Important Improvements in WOMAC Scores After THA and TKA?” by Ghomrawi and colleagues [4], certainly can vie for second-place honors.

The sentence reads: “Most patients having THA and TKA achieved the minimum clinically important difference improvement after surgery” [4].

Only most? I would have thought “nearly every”.

Well, when you scrutinize Table 1 in that paper, you will see that “most” really is the right word, as only 79% of patients who underwent TKA achieved the minimum threshold for improvement in their function and only 65% achieved the minimum threshold improvement in their pain relief.

This is big news!

That at least one-third of all patients who had a TKA by “surgeons [who] are experienced, high-volume joint-replacement specialists from a leading international arthroplasty center” [6] do not attain clinically important improvements in pain after TKA is probably the biggest news Clinical Orthopaedics and Related Research® has ever published.

In fact, the fraction who do not achieve clinically important improvements may be a lot more than one-third, as survey respondents tend to have better outcomes than those who do not respond.

Over the years, I have been citing CORR® in my informed consent discussion with potential TKA patients, but maybe I was using the wrong paper. My go-to source was, “Patient Satisfaction After Total Knee Arthroplasty: Who Is Satisfied And Who Is Not?” by Bourne and colleagues [2]. Their data showed that approximately one-in-five primary TKA patients were not satisfied with the outcome. This paper seems to be considerably bleaker than that!

Needless to say, the work of Watson and Crick got the attention it deserved. I am less sanguine that this powerful result, one that should markedly temper patients’ enthusiasm for surgical solutions, will be as well-known as it should be.

If nothing else, we need to become more adept at identifying these potential failures before they occur. With luck, the work of Fontana and colleagues [3] will help us get there—despite the difficulties inherent in any big data approach [1].