Have you ever wondered what’s the worst academic paper ever published on the subject of neuromonitoring? I ponder this question sometimes as I review the literature because I read so many awful papers. I can’t help but wonder how some of these papers ever get published, and especially in high-profile journals. The problems range from bad research design, to inappropriate statistics, to incomplete methodology, to misinterpretation of findings and outright falsification of data. No one seems to notice. It makes me wonder if I’m the only person out here reading. I know I’m not, but you get the point.
Anyway, neuromonitoring research found its new low in 2010 with the publication of a case report, which I submit is the worst academic paper ever published on the subject of neuromonitoring. At least, it’s the worst one that I’ve ever read. Details are below:
Authors: Jae-Young Hong, MD, Seung-Woo Suh, MD, PhD, Hitesh N. Modi, MS, Chang-Yong Hur, MD, PhD, Hae-Ryong Song, MD, PhD, and Jong Hoon Park, MD, PhD
Journal: Spine (Phila Pa) 35(18): E912-E916, 2010
Description: A 23-year-old neurologically-intact male with a history of spondyloepiphyseal dysplasia presented with severe kyphoscoliosis. Surgical correction and fusion were performed with transcranial MEP monitoring. During the entire procedure, the MEP did not reveal any signs of cord injury. However, lower limb paralysis and paresthesia was observed when the patient awakened. After 2 additional surgical procedures to recover the neurologic deficit, the MEP did not show any signs of cord injury but the patient’s neurologic status had recovered slightly. At postoperative day 8, the neurologic status recovered, and a third operation was performed to fix the long rods. However, there were abnormal amplitudes in both lower limbs but the patient’s neurologic status was almost normal. It is concluded that undesirable events can occur with use of MEP in scoliosis or other spinal surgery. Therefore, we warn the surgeons too heavily rely on the MEP monitoring, and propose a further prospective study as well as use of alternative method that can improve the reliability of single MEP.
I could literally spend hours writing about everything that is wrong with this paper, but we’d both get bored. The fact is, at the end of the day, the most important reason why this is the worst paper ever is that all of the conclusions are based on misinterpretation of data. So, my approach here will be to take you through the data and tell you where the authors went wrong.
So, the procedure was T3-L3 posterior spinal fusion and deformity correction with pedicle screw and rod construct. Prior to dissection, the following baseline MEP data were acquired:
So, this is the baseline study. These are the data to which we compare all future studies in order to provide an ongoing assessment of the patient’s motor function. Here’s the problem: in the image above (A), there were no monitorable MEPs from the legs at baseline. Perhaps there is a blip of a signal in the left leg, but it clearly isn’t of adequate quality for monitoring. From the right leg, the authors monitored 60-Hz noise. So, every judgement made in surgery from here-on is relative to, well, nothing. The authors didn’t see it that way, though. Rather, they saw a 58 μV signal from the left leg, and a 78 μV signal from the right leg. If the authors had any experience in neuromonitoring, then they would have worked to elicit reliable motor evoked potentials from the lower extremities, but I digress.
Following insertion of pedicle screws from T3-L3, the authors performed another test of motor function. Here are the data:
The authors claim that there are no changes in MEPs from baseline after insertion of pedicle screws. I beg to differ. First, the “blip” from the left leg (if it was even real) is now significantly delayed. On top of that, MEPs are now absent from the left arm, likely secondary to malpositioning of that arm. There’s no change in MEPs from the right leg, but they were also absent at baseline. Based on the interpretation of these data as representing no change in spinal cord motor conduction, the authors proceed with correction of the spinal deformity. Following correction, another motor test was performed:
Again, at this point, the authors are still interpreting the data as being consistent with intact spinal cord motor conduction, despite the fact that MEPs remained completely and totally absent from bilateral lower extremities. Guess what happened next! The patient woke with flacid paralysis of the lower extremities, absence of anal tone, absence of deep tendon reflexes, and absence of distal sensations below T6. The authors conclude at this point that neuromonitoring produced a false negative (predicted intact spinal cord conduction in the face of actual paralysis).
So, the authors do a stat CT which shows abnormally placed screws. The patient is rushed back to surgery, the misplaced screws are removed, and a wake-up test reveals that the patient remains paralyzed. At this point, the authors decide to release the rods for fear that they put too much traction on the spinal cord during correction. Here are the data before removal of rods:
There may actually be something there from the legs, for once, but the authors still report no change from baseline. This is a good sign! Following removal of the screws and rods, another test of spinal cord motor function. Here are the data:
Again, the authors interpret the above data as being consistent with baseline (no change). Well, that’s sort-of true since we had no monitorable data from the lower extremities at baseline (Figure A); however, there is a change from data collected before removal of rods (Figure D). Sadly, the authors are still monitoring 60-Hz noise as if it is an MEP signal.
After removal of rods and screws, the patient awakes with some improvement. Specifically, there 1-2/5 strength in the legs, and distal sensations improved from paresthesia to hypesthesia below T6. The Patient was sent to the ICU for recovery.
At postoperative day 8 the patient’s muscle strength had improved to 4/5, and additional surgery was carried out to fix the rods. The contour of the 2 rods was modified to the present curvature to prevent further neurologic injury. During this procedure, the following baseline data were acquired:
So, MEPs are absent from the legs at baseline during this procedure, despite the fact that the patient is relatively neuro-intact (note: this is no different from baseline in the first procedure). After insertion of the rods, a final motor study is conducted and reveals the following data:
So, no change in that there is still no MEPs recorded from the legs. The patient wakes from this surgery with no new neurologic deficits. Thus, the authors interpret this as a neuromonitoring false positive (predict neurologic injury when patient is neuro-intact).
In this case study, the baseline neuromonitoring data were incorrectly interpreted in that MEP signals were identified in the legs, when in fact they were absent. Everything went downhill from there. The patient is lucky to have regained function, but it took 3 surgeries and weeks of recovery. The neuromonitoring system used in this paper looks automated or surgeon-directed. The study highlights the critical need for appropriately-trained, board-certified professional interpreters (AuD, PhD, MD, DC).
Like I said before, I don’t really want to point out all of the little problems that surfaced in this study. It would take a lot of time. With problems this obvious, who needs to nit-pick? This paper is the embodiment of malpractice, and that’s sort-of the irony. When people make egregious mistakes, you almost expect them to try to hide it, but certainly not publish it! That’s how I know that these mistakes were secondary to unfettered ignorance. The authors just didn’t know any better. They didn’t know how to appropriately monitor neural function during scoliosis surgery, and, even if they did, the authors didn’t know how to interpret the neurophysiology data.
I think the worst part of this mess, given that the patient recovered, is that the report got published. It shows that peer-review failed, for one reason or another. Maybe the critiquing parties were also ignorant about monitoring, or maybe they were lazy and didn’t vet the data set. The consequence of the publication is that people cite this paper, when in fact the conclusions are invalid. Indeed, this paper is a wet dream for abstract-skimming neuromonitoring haters, and anyone who cites this paper to validate their own conclusions should be ashamed of themselves. People need to start reading beyond the abstract. When papers like this are used as citations, it creates an impenetrable cycle of misinformation, and the academic arena isn’t the place for this type of nonsense in the age of evidence-based medicine.
The entire paper should be retracted.