Two papers recently published in SPINE argue that neuromonitoring is questionable in ACDF¹ and lumbar fusion² surgery. The papers are written by a group out of UCLA, and these authors have previously published at least one other paper³ questioning the utility of neuromonitoring. Clearly, these spine surgeons don’t like neuromonitoring.
There are some very serious flaws in these papers, and we’ll get to that in a second, but first, I want to talk about the consequences of these papers being published because, believe it or not, if you work in the field of neuromonitoring, these papers affect you directly.
The Consequences of Bad Research
I think sometimes people lose touch with how significant of an impact bad research has on our field at all levels. Consider this: here are 2 big data papers that evaluate neuromonitoring over tens-of-thousands of patients and, on the surface, make a pretty good argument that neuromonitoring has little utility in the very surgical procedures that make up the bulk of the work that we do. These papers get published in widely-read journals and land in the inbox of every spine surgeon in the world.
Major societies like NASS incorporate these papers into their Coverage Recommendations, which are formal position statements about how/when certain methods (e.g., neuromonitoring) should be used and reimbursed. Insurance companies look to NASS Coverage Recommendations to guide their policy, but they also do thorough literature searches and include these papers in their own Coverage Policies, which are formal position statements about how/when neuromonitoring will actually be reimbursed. You can see for yourself here. When everyone believe that neuromonitoring has little utility in ACDF and lumbar fusion surgery, as evidenced by these papers, surgeons use neuromonitoring less frequently and reimbursements drop. I could easily go on-and-on about the consequences of decreasing case volumes and declining reimbursements, but I won’t. Trust me, it affects you directly.
The take-home is that these papers literally cut into our bottom line. When our profits are at risk, our jobs are at risk. Oddly, the major neuromonitoring corporations out there appear disinterested in the topic because I don’t see them doing anything about it. I’ll tell you what I’ve done in a second. What makes matters worse, perhaps, is that I see people from the neuromonitoring community blindly sharing links to these papers on LinkedIn. They even write statements in support of these papers. It makes me SMH and go WTF!
You know, I am 100% in support of taking a good, hard look at our utility in all surgical procedures. We may actually find that we have little utility in some procedures. I don’t know about you, but I don’t want to spend my career working on something where I have no effect, no benefit. What’s the point? Let’s focus on where we’re useful. AMIRITE?!?
So, I just criticized major neuromonitoring companies for appearing disinterested and doing nothing to address this particular problem. That may or may not be fair because sometimes corporations do things behind the scenes. I won’t take it back, but consider the gauntlet thrown down! Let me tell you what I’ve done, and then I’ll get into what’s wrong with these papers.
What I’ve done
First, I teamed up with my friend and colleague, Dr. Adam Doan, and we wrote two letters to the editor in an effort to express our concerns over these papers. The lumbar letter was rejected but sent to the authors for comment; they declined to comment. The cervical letter was accepted4 and sent to the authors for comment; they declined to comment. I’d also like to take a minute just to recognize Dr. Doan’s immense contribution to much of what I’ve written below in the critiques. We worked together on these letters, and i don’t want to take all of the credit.
Second, I went and met with the president of NASS, Dr. F. Todd Wetzel, and told him about our concerns over the papers. Dr. Doan and I then attended a local NASS Town Hall function where we began talking about how neurophysiologists and spine surgeons can work together to look at evidence. Following that conversation, Dr. Wetzel, Doan and I (along with others) wrote a proposal to start a new Section of NASS devoted to neuromonitoring. That application is presently under review.
Third, as an elected member of the ASNM Board of Directors, I went to the board meeting in May and initiated the discussion about how ASNM can get involved in writing letters to the editor when bad research is published. The board was in favor of that idea and the responsibility now falls to the ASNM Research Committee, of which I am also a member. So, we’ll be looking to address bad literature as a society. The Board is definitely working hard to support you!!
So, that’s what I’ve done. In case you’re keeping score, that’s Rich Vogel 3 – Big Corporations 0. Just kidding. So, what’s wrong with these papers? First, just so we’re all on the same page, here is a high-level overview of each paper:
Overview of the Lumbar Fusion Paper
In this retrospective review, the authors used a national database to identify patients that had posterior lumbar fusion (PLF) surgery with and without EMG neuromonitoring for lumbar disorders from years 2007 to 2015. Major conclusions from this study are as follows: 1) The routine use of EMG for monitoring PLF has undergone a highly significant decline since 2009, 2) There are significant regional differences in the use of EMG for PLF, and 3) The risk of neurological complications following PLF in the absence of spinal deformity is low and the routine use of EMG for PLF may not decrease the risk.
Notice how I underlined and bolded the word “use”…? Stay with me….
Overview of the ACDF Paper
In this retrospective study, the authors used the same national database to identify cases of spondylotic myelopathy and/or radiculopathy that underwent ACDF from 2007 to 2014 with and without neuromonitoring. Major conclusions from this study include the following: 1) There was a significant decrease in the use of neuromonitoring for ACDFs from 2007 to 2014, 2) the IONM modalities used for these ACDFs were quite variable, and 3) utilization of neuromonitoring varies regionally.
Notice how I underlined and bolded the word “use” and “utilization”…? Stay with me….
Critique of Both
OK. Let’s first talk about the fatal flaw that is common to both the lumber fusion and the ACDF paper. In both papers, the authors’ use a national database called PearlDiver, which is a commercially-available database containing de-identified patient information from Humana. They used this database to analyze national trends in the utilization of neuromonitoring. The problem is that this database cannot answer questions about utilization because data are based on claims paid or adjusted (e.g., for neuromonitoring). PearlDiver does not contain information about claims denied payment or claims not submitted for payment (personal communication confirmed with PearlDiver). In other words, data extracted from PearlDiver can only tell us how many times Humana agreed to pay for neuromonitoring during ACDF or lumbar fusion surgery, and it cannot tell us how many cases were actually monitored. Are you with me?
This observation alone completely invalidates every single statement that the authors make regarding utilization of neuromonitoring. That’s why I underlined and bolded the words “use” and “utilization” above!
Now….I could drop the mic right here, but why stop now? Let’s go through the exercise and see what else is wrong with these papers. Let’s start with the lumbar fusion paper.
Critique of the Lumbar Fusion Paper
At one point in the paper, the authors state, “There was a steady increase in the use of EMG from 14.9% in 2007 to a peak of 28.7% in 2009. This was followed by a steady decrease to 21.9% in 2015 (p < 0.0001) (Figure 2).” Unfortunately, this statement is discordant with the data presented in Figure 2 of this same study. I wish I could reprint the figure, but you should really pull the paper if you have access. According to that figure, and taking into account that PearlDiver describes reimbursement and not usage, there was indeed a steady increase in the reimbursement of EMG from 2007 to 2009; however, the reimbursement of EMG after 2009 appears to fluctuate (not decrease).
Assuming that the authors analyzed data only between 2009 and 2015 to draw that conclusion (excluding data from 2007 and 2008), the Chi Square test suggests that there is a highly significant difference somewhere within the range of years 2009-2015, but the test does not (and cannot) find “a steady decrease”, nor can the test identify a “significant decrease” across a range of years. The Chi Square test can only tell us whether there were 2 years in that range that differed significantly from each other. In order to identify a significant trend in proportional data across a series of years, the authors would need to use a completely different statistical test, such as Spearman’s Rank Correlation or one of the available regression models (logit/probit). In any case, the authors make a statement that is simply incorrect. This statement is not supported by the Figure in which the data are illustrated, and the statement cannot be supported by the statistical test that the authors elected to employ.
As further evidence of the equivocal nature of the authors’ results, other studies examining the utilization of IONM during the same interval of years have reported very different results.5-7 For example, James and colleagues5 examined 443,194 spinal procedures from 2007-2011 using the AHRQ NIS database and found that the percentage of lumbar spine surgeries in which neuromonitoring was used increased rather dramatically from 2007-2011 as follows: 0.6%, 7.2%, 7.5%, 11.8% and 12.6%. In this longitudinal study, the use of neuromonitoring increased in similar fashion irrespective of spine location (i.e., cervical, thoracic, thoracolumbar and lumbar). Also, there was a rather dramatic increase in the use of IONM in lumbar spine surgery during some of the same years in which Ajiboye et al argue that the “use” was on a “steady decline”. Thus, a much larger national database which actually tracks usage of neuromonitoring5 returned very different results from those reported by this lumbar fusion study, and this further supports the argument that the results reported Ajiboye and colleagues are erroneous.
Next, commercial Humana patients do not constitute a sample that is representative of neuromonitoring nationally. PearlDiver returned 9,957 patients undergoing posterior lumbar fusion during an eight-year period (2007-2015). Of those patients, 2,494 received neuromonitoring, yielding an annual average of 311 patients. I’ll bet even the smallest neuromonitoring companies out there do more than 300 posterior lumbar fusions each year. So, the sample cannot be representative of the population.
Want more? Let’s look at the rate of postoperative deficits. The authors found rates that were roughly equivalent between those with neuromonitoring (1.36%) and those without (1.34%), but they only included one test modality: EMG. In doing so, they neglected tests that are actually designed to detect neurological deficits. Although triggered EMG is useful to test the approximation of instrumentation to neural tissue (i.e., medial breach), it is not considered a reliable predictor of postoperative deficits (PPV ≈ 10%). As such, it is rarely used independently of additional test modalities (SSEP and MEP).
Also, one could argue that medially-breached screws don’t always result in deficits, and one could also argue that many deficits in lumbar surgery aren’t secondary to instrumentation. We also have no way of knowing if the sample is formed by a selection bias, such as neuromonitoring only being used for higher risk patients/surgeries, which is often the case. I am not aware of any large-scale studies that examine postoperative deficits with unimodal versus multi-modal monitoring in lumbar spine surgery.8
Failure to publish data of this nature, which is readily available for hundreds of thousands of surgeries performed by outsourced neuromonitoring providers throughout the USA, is a long-standing problem in the field of intraoperative neurophysiology and now it’s starting to catch up to us. You can take that to the bank! This provides a really good segue to looking at problems with the ACDF paper.
Critique of the ACDF Paper
You will recall that a central conclusion of this paper was that there was a significant decrease in the use (actually reimbursement) of neuromonitoring between 2007 and 2014. I think we all understand that the challenges we face in seeking reimbursement for neuromonitoring are significant; particularly with the erosion of concurrency which began in 20139. Additionally, as I just mentioned, our profession’s history of under-publishing results, despite the preponderance of data, has now introduced new obstacles as payors, like Humana, frequently change their Medical Coverage Policies (MCP) to limit reimbursement for our services.
Neuromonitoring is reimbursed by diagnosis, not by procedure, and a critical factor that Ajiboye and colleagues failed to consider is that Humana changed their MCP several times regarding reimbursement for neuromonitoring during this paper’s analysis period. Ooops!
Take a look at this: In 2011, Humana added new language to their MCP for neuromonitoring, which states:
Decompression or discectomy for disc herniation and laminectomy and decompression for stenosis should not require intraoperative neurological monitoring because monitoring is not likely to alter the surgical outcome for these conditions.10
The addition of this new language clearly laid the foundation for decreasing neuromonitoring reimbursements for cervical discectomies. Thus, it is entirely possible, actually probable, that the reduction in reimbursements for ACDF that the authors observed in 2012 is directly related to the fact that Humana changed their MCP in the latter half of 2011. Again, this is reimbursement, not utilization as the authors suggest.
One More Critique of Both
A final comment that I’ll make regarding both papers is related to the use of big data and the rather ubiquitous problem of heterogeneity. Many people think of meta-analyses and systematic reviews as being near the apex of the research pyramid, but that’s not always true. In the neuromonitoring profession, they are rarely used to draw credible conclusions because there is too much variability in patient populations, modality selection (e.g., EMG, SSEP, MEP), methodology, alert criteria and types of post-alert interventions, just to name a few common variables that cannot be controlled across large samples.
Being a younger and less-regulated field, our methodology isn’t as uniform as one might expect of a more established profession. As such, we still lack consistent guidelines, secondary education programs, uniform board certification, and licensure. Similarly, we work in the context of significant variability in anesthesia regimens and ways in which surgeons react to our information. Although Ajiboye and colleagues1-2 mention some of these limitations in their studies, they don’t delve deeply into these critical factors that affect how we view the sensitivity and specificity of neuromonitoring data – both of which guide decisions about utilization and reimbursement.
In our letters to the Editor of SPINE, we recommended as a course of action that the Ajiboye and colleagues retract all statements related to the utilization of neuromonitoring, and instead provide an examination of 1) the litany of variables (listed in the paragraph above) that preclude one from making substantive statements about the effectiveness of neuromonitoring in reducing postoperative neurological deficits, and 2) the myriad reasons why reimbursement for neuromonitoring is on the decline at a time when utilization is clearly on the rise. Either one of these would actually be a useful contribution to the literature.
- Ajiboye RM, D’Oro A, Ashana AO, Buerba RA, Lord EL, Buser Z, Wang JC, Pourtaheri S. Routine Use of Intraoperative Neuromonitoring During ACDFs for the Treatment of Spondylotic Myelopathy and Radiculopathy Is Questionable: A Review of 15,395 Cases. Spine (Phila Pa 1976). 2017 Jan 1;42(1):14-19.
- Ajiboye RM, et al. The Utility of Intraoperative Neuromonitoring for Lumbar Pedicle Screw Placement is Questionable: A Review of 9957 Cases. Spine (Phila Pa 1976). 2016 Nov 15. [Epub ahead of print].
- Ajiboye RM, Zoller SD, Sharma A, Mosich GM, Drysch A, Li J, Reza T, Pourtaheri S. Intraoperative Neuromonitoring for Anterior Cervical Spine Surgery: What Is the Evidence? Spine (Phila Pa 1976). 2017 Mar 15;42(6):385-393.
- Doan A, Vogel R. To The Editor. Spine (Phila Pa 1976). 2017 May 22. [Epub ahead of print] PubMed PMID: 28538592.
- James WS, Rughani AI, Dumont TM. A socioeconomic analysis of intraoperative neurophysiological monitoring during spine surgery: national use, regional variation, and patient outcomes. Neurosurg Focus. 2014 Nov;37(5):E10.
- Ney JP, van der Goes DN, Nuwer MR. Does intraoperative neurophysiologic monitoring matter in noncomplex spine surgeries? Neurology. 2015 Dec 15;85(24):2151-8.
- Vadivelu S, Sivaganesan A, Patel AJ, Agadi S, Schmidt RJ, Mani P, Jea A. Practice trends in the utilization of intraoperative neurophysiological monitoring in pediatric neurosurgery as a function of complication rate, and patient-, surgeon-, and procedure-related factors. World Neurosurg. 2014 Mar-Apr;81(3-4):617-23.
- Fehlings MG, Brodke DS, Norvell DC, Dettori JR. The evidence for intraoperative neurophysiological monitoring in spine surgery: does it make a difference? Spine (Phila Pa 1976). 2010 Apr 20;35(9 Suppl):S37-46.
- Ney JP. Changes to CMS Reimbursement Rules for Intraoperative Neurophysiological Monitoring: Implications for Telemedicine. Telemed J E Health. 2013 Oct; 19(10): 791–793.
- Humana. Medical coverage policy on intra-operative neurological monitoring (Policy Number: CLPD-0336-005) [Humana web site]. July 28, 2011. Available at: http://apps.humana.com/tad/tad_new/home.aspx. Accessed and downloaded July 28, 2011.