Discussion: Appropriateness of the Use of Magnetic Resonance Imaging in the Diagnosis and Treatment of Wrist Soft Tissue Injury

    loading  Checking for direct PDF access through Ovid


In this review of electronic medical records at the University of Michigan, Michelotti and colleagues1 found that a wrist magnetic resonance imaging scan ordered by a hand surgeon is more likely to alter the diagnosis and/or treatment recommendation than one ordered by a non–hand surgeon. One interpretation of this finding is that hand surgeons use magnetic resonance imaging to answer more specific diagnostic questions (i.e., they use physical examination and plain film to narrow the differential diagnosis, which increases the pretest probability of disease, which in turn improves the diagnostic performance of magnetic resonance imaging), whereas non–hand surgeons might be taking more of a “shotgun” approach, a strategy that can be more misleading than helpful.
It is crucial that we all—including hand surgeons—understand the statistics underlying this shortcoming. For the sake of illustration, let us use data from a recent study of patients with previous wrist trauma and ulnar-side wrist pain, setting the sensitivity and specificity of 1.5-T magnetic resonance imaging for soft-tissue injury at 0.73 and 0.86, respectively.2 Now let us say that soft-tissue injury is the cause of wrist pain in 15 percent of cases. In a population of 100 patients, magnetic resonance imaging would produce 11 true-positive results (15 × 0.73) and 12 false-positive results (85 × 0.14), for a positive predictive value of 48 percent. In other words, in this scenario, a patient with a “positive” magnetic resonance imaging scan is slightly more likely not to have a “real” soft-tissue injury. Without an appreciation of this relationship between prevalence and predictive value, magnetic resonance imaging could easily lead to misdiagnosis and overtreatment.
The authors observed that “[t]he only predictor of surgery following magnetic resonance imaging was if the study was ordered by a hand surgeon (26 percent versus 8 percent; p = 0.006),” and argued that this was likely true for the reason discussed above. Indeed, if the prevalence in the above scenario were 50 percent (after the hand surgeon winnowed the cohort), the positive predictive value for soft-tissue injury jumps to 84 percent. However, an alternative interpretation is that hand surgeons are more likely to look for and find diagnoses that support operative treatment. In this light, the suggestion that hand surgeons should be the ones who order magnetic resonance imaging scans, make final diagnoses, and offer treatment might lead to more resource use and potential iatrogenic harm.
This suggestion also ignores the other side of the statistical coin. In the low-prevalence environment of 5 percent (as might be seen in a primary care physician’s office), the negative predictive value for magnetic resonance imaging would be 98 percent. If the goal were to rule out a discrete, treatable abnormality and focus on palliation and resiliency, a negative magnetic resonance imaging result could thus be quite valuable, especially if it helps get an employee back to work. The caveat is that humans have trouble ignoring the “diagnoses” written on a magnetic resonance imaging report. “Triangular fibrocartilage complex tear,” “extensor carpi ulnaris degeneration,” “ganglion cyst”—these words change the way a person thinks about their wrist, with consequences on their symptoms and limitations (i.e., their health).
Thus, the real question is whether a magnetic resonance imaging scan is likely to have a positive or negative impact on health. The answer must account for two realities. First, nonspecific wrist pain is extremely common. It is estimated that half of all wrist pains have no clear pathophysiologic explanation,3 and even diagnostic arthroscopy cannot improve this uncertainty.4 Second, incidental findings on wrist magnetic resonance imaging are extremely common. For instance, half of all asymptomatic wrists have a ganglion cyst.

Related Topics

    loading  Loading Related Articles