From Bench to Bedside: Improving the Magic 8 Ball—Prognostication for Osteoarthritis

Most of us will remember the Magic 8 ball from our youths: Shake the ball and get an “accurate,” often humorous, answer to your no-doubt silly question. It was a bit like a fortune cookie roulette wheel. Medicine has been working on a better version of the Magic 8 ball, in any number of disciplines, for decades. Indeed, we have “prognostic” staging systems for all kinds of malignancies [1], including arthritis, in almost any joint you can think of—except those in the middle ear (which actually exist, but are quantified by degree of hearing loss [2]).

I am not disparaging those systems and have even created one myself. They can be helpful in terms of guiding treatment, patient counseling, and comparative research validity or data binning, and we use them all the time. But these systems simply quantify what has already occurred, guiding treatment and “predicting” outcomes by grading where the patient is now along a spectrum of disease badness. A patient with stage III disease will more likely need surgery than a patient with stage I disease, sure, but which patients with stage I will progress to stage III? What we are really pursuing is a Magic 8 ball that works for more than youthful hinjix and entertainment: a Palantír, or crystal ball.

We recently made a substantial step forward in this regard for predicting progression of osteoarthritis (OA), which affects hundreds of millions of people and is a leading cause of disability worldwide [5]. Zhou and colleagues [6] very recently reported on a panel of 15 serum proteomic biomarkers that can distinguish knee OA progressors from nonprogressors. The specifics of their study are beyond both the scope of this column and my ability to explain them. But that panel of biomarkers turns out to be a pretty good predictive tool to anticipate whether a patient will have progressive OA. Its area under the receiver operating characteristic curve was 0.73, which is a substantive improvement over prediction using standard clinicoradiologic indicators (0.59) or the urinary C-terminal cross-linked telopeptide of type II collagen (0.58), which is the best standalone biomarker [4] (remember that 0.5 is analogous to a random guess and 1.0 is a perfect prediction tool).

There are, of course, limitations to that study. Time is a critical factor and huge barrier in research that seeks to describe patients’ prognoses with the hope of eventually intervening to improve them. The research team followed patients for just three or four years, which might seem like a long time in orthopaedic clinical research, but in the context of a chronic disease like OA that can take decades to evolve, this is on the short side. Additionally, most patients already had moderate-to-severe OA at baseline, and the mean participant age was 65.8 years [6]. This is why the authors attempted to distinguish progressors from nonprogressors rather than, for example, OA developers versus nondevelopers. Is osteoarthritis already too advanced in these cohorts for some effective future therapy or intervention to be effective? A critical finding, which assuages some of these concerns, is that their model performed better when patients with more-severe OA were removed.

Similar efforts in post-traumatic osteoarthritis (PTOA) research have attempted to identify relevant serum or synovial fluid markers present shortly after injury [3]. This is a compelling target population because the affected patients are often younger, and PTOA, if and when it develops, typically occurs more rapidly than age-related OA. The challenge here is separating causal biomarkers that might actually be part of the pathogenesis of rapidly developing PTOA from those that merely are associated with the condition (such as those that might be present after a severe intra-articular fracture in all patients). Proteins that cause the condition or contribute to it are potential targets for intervention (perhaps with antagonists or blocking antibodies), while the others are, at most, innocent bystanders. This represents the greatest “chicken or egg” challenge of both OA and PTOA research.

The next step is further validation. Have Zhou and collaborators [6] actually discovered a crystal ball? This should happen concurrent with, or followed by, testing in patients with earlier stage OA or no OA at all. The ultimate goal here, however, is not just to predict OA; rather, by identifying patients who do not yet have OA (or patients with mild-to-moderate OA who are very likely to have rapidly progressive disease), we can more effectively design and power interventional trials to prevent, stop, or reverse OA.

At the end of the day, the real question is: If we know the future, can we change it?