Editor’s Spotlight/Take 5: Routine MRI Among Patients With a Suspected Scaphoid Fracture Risks Overdiagnosis

At first, I thought it was me. This study [5] seemed to say that it’s okay to not look too hard for scaphoid fractures, and I found the idea stressful, if not frankly transgressive. After all, in residency, I learned that if a patient presents with tenderness over the anatomic snuffbox (such a quaint term), but the radiographs are negative, the patient gets a cast and another plain film in a couple weeks. Important, I was taught, because if I blew that diagnosis, I could leave someone with a lifetime of wrist pain and disability. Fast forward to the modern era. That same patient in typical practice probably gets an MRI now. Current residents, soon to be attendings—fed wholly or in part on a diet of pellets from Orthobullets—are likely to order that MRI, since that source says that both the sensitivity and, importantly, the specificity, approach 100% for occult fractures when imaged with MRI [28]. And now, this, in the current issue of Clinical Orthopaedics and Related Research®: “Routine MRI of suspected scaphoid fractures carries a notable risk of overdiagnosis and potential overtreatment” [5]. Quitting something we thought was right is hard. It doesn’t happen as often as it should in medicine (and surgery), and when it happens, it often happens long after the evidence is solid. It takes nearly two decades for good evidence to result in better practice [27]. Looks like we’re running true to form on knee arthroscopy for degenerative meniscal tears and related diagnoses—robust studies have been around for nearly that long, but we continue to perform that operation in great numbers [22]. Same goes for many other common orthopaedic procedures [21]. Speaking personally, I understand surgeons’ reluctance to change. When I began my practice, same-day bilateral TKA was entirely normative, and I probably did about a third of my TKAs that way. In time, evidence began to accumulate that doing this was not as safe as we thought [30], but those studies seemed at odds with my own observations; my patients seemed happy with me, I thought. The skeptic might retort “except those who needlessly died,” which would’ve been a good point and one that would be easy to miss in that context. After all, the percentages are so small that they could elude observation even by an attentive surgeon, while still resulting in enough excess harm that by any measure would be worth caring about. So it wasn’t the first study questioning the safety of same-day bilateral TKA that convinced me to stop doing them, and it probably wasn’t the second. Realizing that I probably caused needless harm was hard to accept. It took me some time. For that reason, I do understand the struggle faced by surgeons still performing meniscal debridements in middle adulthood [22], procedures for subacromial “impingement” [29], and surgery for low back pain [4], among numerous other discredited operations [21]. I just hope they’re accepted soon. The current study in CORR®, “Routine MRI Among Patients With a Suspected Scaphoid Fracture Risks Overdiagnosis” [5], makes a good case that if we stop chasing occult scaphoid fractures, we will prevent many patients from getting overtreated. Nearly 70% of the MRI findings they found in this context were categorized as “distracting and potentially misleading,” and only 2% of those suspected of injury actually had definite fractures. The authors point out that the Bayes theorem holds that with the real risk being so low, and the frequency of misleading findings being so great, the potential for unnecessary surgery may be greater than the benefit to the few patients whose fractures are correctly identified. At the very least, their suggestion of hosting a thoughtful conversation with each patient about the potential harms not just of treatment strategies, but also diagnostic approaches, seems entirely worth adopting. Although I don’t treat patients with this injury (I’m a hip and knee guy) and most readers here probably don’t either, the issues this paper raises about overdiagnosis and overtreatment apply to all of us, and therefore deserve all of our attention. Please join me as I go behind the discovery to talk about these issues and others with David Ring MD, PhD, author of “Routine MRI Among Patients With a Suspected Scaphoid Fracture Risks Overdiagnosis” in the Take 5 interview that follows. Take 5 Interview with David Ring MD, PhD, author of “Routine MRI Among Patients With a Suspected Scaphoid Fracture Risks Overdiagnosis” Seth S. Leopold MD:Congratulations on this mind-opening study. We’ve covered overtreatment a great deal in these pages [23, 24], and it often focuses on unnecessary surgery, which is, by definition, harmful. But in the context of the suspected scaphoid fracture, overtreatment seems most likely to be just casting a patient (or casting a patient for a little longer). How big of a deal is this, really? David Ring MD, PhD: Let’s broaden the concept of harm to include not just adverse outcomes and iatrogenic harms, but also psychological and financial harms. Considering Oxford’s use of routine MRI among 258 people with a suspected scaphoid fracture referred for MRI [7], 45 did not attend (inconvenience) and 5 did not tolerate the scan (because of distress). Here, we see some of the psychological harms: Do any of these individuals carry some guilt or misgivings? Among the 208 patients who had an MRI, 13 (6%) were diagnosed with scaphoid “contusions” and 27 (13%) with “fractures.” If the distinction between contusion and fracture is not reliable or accurate, that could lead to a notable potential for overdiagnosis or underdiagnosis. It’s difficult to know the accuracy of MRI because we don’t have a consensus reference standard for true scaphoid fractures. But suppose there are 5% false positives (95% specific) on MRI in this setting. Bayes theorem tells us that only 74% of MRI-diagnosed fractures are true fractures in a setting of 13% true scaphoid fractures. That means 1 in 4 people are overdiagnosed and overtreated. Even if MRI is 98% or 99% specific as some claim, 1 in 10 people are misdiagnosed in this setting. To paint a “worst-case” circumstance, let’s plug in the Oxford numbers that 67% of scaphoid signal abnormalities were fractures and 33% were not. In that circumstance, scaphoid signal abnormalities will indicate a true fracture only 23% of the time (and 77% will be an overdiagnosis). Or, in an even worse “worst-case” scenario, if you plug in the numbers from our study—69% potentially distracting signal abnormalities and 31% definite or possible fractures—that leads to a 6% probability of the MRI diagnosing a true fracture. This is not just a thought experiment. Some studies interpret signal abnormalities as contusions that other studies interpret as fractures. And the diagnosis of nonfracture signal abnormalities as fractures or potential fractures is probably even more common in daily practice, where we often see the “cannot rule out” wording in radiology reports. Continuing with the Oxford data, let’s look at the 27 patients diagnosed with a radiographically invisible scaphoid fracture. Those patients received cast immobilization and were monitored radiographically. Two of 27 (7%) were diagnosed with delayed unions and accepted an offer of operative treatment. Given the unreliability of radiological diagnosis of union [8, 13, 36] and the lack of knowledge regarding the natural history of a radiographic diagnosis of “delayed union,” it’s quite possible that these patients (1 in 14) were overdiagnosed and overtreated. The potential harms associated with self-management of a radiographically invisible scaphoid fracture seem low for several reasons. Most scaphoid nonunions are associated with fracture displacement [1, 3]. Relatively short periods of protection may be sufficient for nondisplaced fractures [12]. Invisible fractures (zero displacement) might heal without immobilization. It seems plausible, even likely, that the rate of undiagnosed, untreated, nondisplaced scaphoid fractures among people with sore wrists after a fall (likely common) is greater than the rate of diagnosed scaphoid nonunion (uncommon), which would suggest that many nondisplaced scaphoid fractures heal without treatment. One of the key areas of contention regarding a suspected scaphoid fracture seems to be radiograph findings during the first 6 months. It’s not clear that we can accurately diagnosis fractures that will go on to nonunion. And we also don’t know the natural history of stable persistent radiographic fracture lines (so-called “fibrous union”). Among the 115 nonunions in the series of Herbert and Fisher [14], 33% (38) were deemed fibrous. They chose to operate on 35 of these fibrous unions; most just got a screw (no debridement of the fibrous union and no bone graft) and all but 2 healed. I wonder how many of those would have healed without surgery. I also wonder what would have happened if they did not heal. The natural history of untreated “fibrous unions” of the scaphoid waist is unknown and might be benign. The relatively good results of earlier surgery for “delayed union” [32] and of screw fixation without graft for “fibrous union” may represent overdiagnosis and overtreatment. Worried surgeons may be operating on fractures with a visible fracture line that were healing. Dr. Leopold:Okay, I get it, but help me with the stress I mentioned in the essay at the top here. How do I balance your answer to the previous question against the risk of a devastating event—say a small chance of a young patient turning up with a nonunion and collapse from a missed fracture (a hand surgery colleague of mine says he sees one or more of these each year in clinic), weighed against the risk of a bunch of people wearing casts a bit longer? “The notable risk of symptomatic or problematic nonunion,” as you put it in your paper, is really, really bad, right? Dr. Ring: Posttraumatic arthritis associated with a scaphoid nonunion is varied and often well accommodated [15, 18]. We sometimes meet people decades after a fracture, remembered or not, who experienced no symptoms until a few months prior [15, 18, 26]. Presumably, some of those patients never seek care. When facing decisions about potential harms and potential benefits, should a person with a sore wrist decide, or should the clinician decide? The answer to the question, “what probability of posttraumatic wrist arthritis would you be willing to accept for the tradeoff of being able to self-manage your sore wrist?” is likely to vary from patient to patient. To what degree should the clinician’s worries (including worries about being sued) play a role? There is a relative consensus that the risk of symptomatic nonunion with arthritis in the setting of a suspected scaphoid fracture is less than 1%. Many patients would likely consider that low risk acceptable. That’s been the experience in Fife, Scotland, where they have initiated a strategy of patient self-management of suspected scaphoid fractures. Among 221 patients with self-managed suspected scaphoid fractures, only 9% (19 of 221 patients) were called in after specialist review of the radiographs identified a potential problem (7 diagnosed with a scaphoid fracture, 8 with another injury, and 4 as normal). And only 6% (13 of 221) sought more care on their own, with no injury identified in any of those 13 patients. Dr. Leopold:You’ve spent time thinking about overdiagnosis and overtreatment. Can you generalize from what you learned in this study more broadly for our non–hand surgeon readers? What analogies do you see from this study that are playing out in orthopaedics more generally, and how much of a problem do we have? Dr. Ring: The problem is well-defined by Nobel prize winners such as Kahneman [16] and Thaler and Sunstein [35]: Humans overvalue and overtrust the autopilot mind. We often inadequately use our critical and analytical mind. Humans benefit from anticipating that the automatic and intuitive thoughts that arise from heuristics (mental shortcuts) are often inaccurate. Humans also benefit from doing the work to set up, and practice using, systems to catch those errors before they cause harm. There are some principles that musculoskeletal specialists can use to help catch errors. First, tests have more potential for harm when the clinician is puzzled than when the test is confirming a likely diagnosis. This is Bayes’s emphasis on prior information. Second, we don’t need to link all symptoms to a pathophysiology. If there is a low pretest probability of a pathophysiology that would benefit from treatment, then we are probably better off not ordering the test. Any important but low-risk pathophysiology will be discovered over time, and the delay usually causes little or no harm. There are many, many pathophysiologies for which the diagnosis will not improve our health appreciably or might even harm our health. These can go safely and healthfully undetected. Third, a lot of pathophysiology is either asymptomatic or well accommodated. We should be prepared for incidental test findings. The most healthful strategy is to ignore them. Given the many musculoskeletal pathophysiologies that are associated with self-limiting symptoms such as many enthesopathies and tendinopathies, in combination with the numerous pathophysiologies of senescence that are near universal with age [2, 19, 34], the decision about whether to ignore incidental test findings and accommodate the sensations in one’s aging body is an everyday experience for musculoskeletal specialists and their patients. These principles often are bypassed or ignored in the biomedical paradigm in which an attempt is made to link every symptom to some pathophysiology and then treat that pathophysiology. This “find it and fix it” mentality greatly increases the risk of overdiagnosis and overtreatment. Dr. Leopold:What would it take for us to decrease the harms caused by overdiagnosis and overtreatment to patients with musculoskeletal problems? Articles in journals are not doing the trick, it seems… Dr. Ring: One thing to learn from thinking deeply about suspected scaphoid fractures is that the possibility of underdiagnosis and undertreatment cannot be eliminated. Patients and physicians alike must therefore choose the levels of underdiagnosis and overdiagnosis and undertreatment and overtreatment that are consistent with what matters most to them (their values), unclouded by common misconceptions and misinterpretations. The importance of critical/analytical thinking cannot be overstated because we often misperceive and misinterpret our body’s sensations [20, 25] as well as the available options for good health [33]. With misperceptions and misinterpretations adequately identified and reoriented—a step that is often skipped—people can then match their foundational priorities and values to the probability of the variety of harms associated with specific diagnostic strategies. Cultural change is needed. Humans need to learn and practice how to operate and make best use of their minds. Appreciate your mind and value it, but do not trust it. And there is evidence that can give us hope: The more you practice analytical thought, including Bayesian thinking, the more intuitive and automatic it becomes [17]. This is an endeavor for all humans, including musculoskeletal specialists. Dr. Leopold:Back to scaphoids. If the risk of overdiagnosis is high, why not just make better criteria? Define what is definite, and what is likely to be misleading? You did this in the paper; would using these criteria lead to fewer misdiagnoses? You mention using a clinical prediction rule; how can/should this approach factor in? Dr. Ring: In my view, Bayes’s revelation that even very accurate tests can be misleading when the pretest odds of a pathophysiology being present are low suggests that we are not going to solve this problem with better imaging or better interpretation of imaging. Additional sources of uncertainty include shortcomings and inconsistencies in imaging technology and the images produced, the absence of a consensus reference standard by which to analyze test accuracy, the prevalence of nonfracture signal abnormalities, and the imperfect reliability of interpretations [31]. Surgeons can anticipate advances in image technology and quality and work toward more discerning and reliable interpretations of the imaging as we and others have done [6]. But to arrive at a better balance of potential harms and benefits, I think we may need to first accept the unresolvable uncertainty, and then try to use imaging only in settings where the pretest odds are acceptably high and diagnosis of a fracture would alter the treatment strategy. Clinical prediction rules use symptoms, signs, and test results to estimate the probability of disease. These show some promise for increasing the pretest odds of a true fracture in the setting of a suspected scaphoid fracture [10]. Let’s look further into the part about diagnosis altering the treatment strategy. Given the evidence that a nondisplaced fracture doesn’t benefit from more than 4 weeks of immobilization, and if the wrist immobilization can take the form of a removable brace, then simply bracing the wrist for a month can be an alternative to testing for a true fracture. That might be an acceptable middle ground for many people—treat my wrist as if it has a fracture, just in case. People who want to be brace-free sooner, and who ideally have a relatively high pretest odds of a fracture, might consider additional imaging such as MRI. People who can accept a less than 1 in 100 possibility of symptomatic nonunion with arthritis can return immediately to low-risk activities without a brace and without additional imaging. A person who wants to return to a risky activity promptly might benefit from the high negative predictive value of an MRI, in large part because of the low prevalence of true fractures among suspected scaphoid fractures. But it may be, as in Fife, that most of us are happy to simply take it easy for a while, and nothing more is needed for a sore wrist. Another consideration is that, given the relationships between discomfort with uncertainty, pain anxiety, and greater symptom intensity [11], people most at risk of harm from overdiagnosis and overtreatment may be those with relatively low risk. Discomfort with uncertainty undermines health in circumstances of both underdiagnosis and overdiagnosis and can be addressed as a part of the illness [9].David Ring MD, PhD