In 1989, a group of sixty-seven asymptomatic individuals with no history of back pain or sciatica underwent magnetic resonance scans of the lumbar spine. In a landmark 1990 JBJS study, Boden et al. reported that three neuroradiologists who had no clinical knowledge of the patients interpreted the images as being substantially abnormal in 28% of the cohort (19 individuals). More specifically, a herniated nucleus pulposus was identified in 24 % of these asymptomatic subjects. These “magnetic-resonance positive” findings were more prevalent in older subjects; abnormal MRI findings were identified in 57% of those aged 60 to 80 years.
Boden et al. concluded that so many MRI findings of substantial abnormalities in asymptomatic people “emphasized the dangers of predicating a decision to operate on the basis of diagnostic tests—even when a state-of-the-art modality is used—without precise correlation with clinical signs and symptoms.”
However, despite the findings of Boden et al., during the last five years of the 1990s, Medicare claims showed a 40% increase in spine-surgery rates, a 70% increase in fusion-surgery rates, and a two-fold increase in use of spinal implants. Although spine-fusion surgery has a well-established role in treating certain spinal diseases, a 2007 systematic review of several randomized trials indicated that the benefits of fusion surgery were limited when treating degenerative lumbar discs with back pain alone. This review suggested the need for more thorough selection of surgical candidates, which was a caution also implied by Boden et al.
Although the three neuroradiologists in the Boden et al. study largely agreed on the absence or presence of abnormal findings on the MRIs, in 2014 Fu et al. reported on the interrater and intrarater agreements by four reviewers of MRI findings from the lumbar spine of 75 subjects. Even though this study used standardized evaluation criteria, there was significant variability in both interrater and intrarater agreement among the reviewers. As the Boden et al. study did 25 years ago, this study demonstrated the diagnostic limitations of MRI interpretation for lumbar spinal diseases.
In 2001, JBJS published a paper by Borenstein et al. that was a seven-year follow-up study among the same asymptomatic subjects studied by Boden et al. Borenstein et al. found that the original 1989 scans of the lumbar spine were not predictive of the future development or duration of low back pain. This led Borenstein et al. to conclude—as Boden et al. did—that “clinical correlation is essential to determine the importance of abnormalities on magnetic resonance images.”
Many important subsequent studies were inspired by the original findings of Boden et al. in JBJS. Most of them emphasize that for lumbar-spine diagnoses, an MRI is only one (albeit important) piece of data; that interpretation of MRIs is variable; and that all imaging information must be correlated to the specific patient’s clinical condition.
Several studies and national surveys indicate that approximately a quarter of US adults report having had back pain during the past 3 months, making this a common clinical complaint. But the findings of Boden, et al. and subsequent studies remind us that surgery is not always the appropriate treatment.
Daisuke Togawa, MD, PhD
JBJS Deputy Editor