Article Text

Download PDFPDF

Should neurointerventional fellowship training be suspended indefinitely?
  1. David Fiorella1,
  2. Joshua A Hirsch2,
  3. Henry H Woo1,
  4. Peter A Rasmussen3,
  5. Muhammad Shazam Hussain3,
  6. Ferdinand K Hui3,
  7. Donald Frei4,
  8. Phil M Meyers5,
  9. Pascal Jabbour6,
  10. L Fernando Gonzalez6,
  11. J Mocco7,
  12. Aquilla Turk8,
  13. Raymond D Turner9,
  14. Adam S Arthur10,
  15. Rishi Gupta11,
  16. Harry J Cloft12
  1. 1Cerebrovascular Center, Department of Neurosurgery, State University of New York at Stony Brook, Stony Brook University Medical Center, Stony Brook, New York, USA
  2. 2Neuroendovascular Program, Massachusetts General Hospital, Boston, Massachusetts, USA
  3. 3Cerebrovascular Center, Cleveland Clinic, Cleveland, Ohio, USA
  4. 4Department of Interventional Neuroradiology, Radiology Imaging Associates, Swedish Medical Center, Denver, Colorado, USA
  5. 5Departments of Radiology and Neurological Surgery, Columbia University, College of Physicians and Surgeons, NYC, New York, USA
  6. 6Department of Neurosurgery, Thomas Jefferson University Hospital, Philadelphia, Pennsylvania, USA
  7. 7Department of Neurosurgery, Vanderbilt University, Nashville, Tennessee, USA
  8. 8Cerebrovascular Center, Department of Radiology, Medical University of South Carolina, Charleston, South Carolina, USA
  9. 9Cerebrovascular Center, Department of Neurosciences, Medical University of South Carolina, Charleston, South Carolina, USA
  10. 10Department of Neurosurgery, Semmes-Murphey Neurologic and Spine Institute, Memphis, Tennessee, USA
  11. 11Department of Neurology, Marcus Stroke and Neuroscience Center, Grady Memorial Hospital Atlanta, GA, USA
  12. 12Department of Radiology, Mayo Clinic, Rochester, MN, USA
  1. Correspondence to Dr David Fiorella, Cerebrovascular Center, Department of Neurosurgery, State University of New York at Stony Brook, Stony Brook University Medical Center, Health Sciences Center 080, Stony Brook, NY 11794, USA; david.fiorella{at}

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

To bring about destruction by overcrowding, mass starvation, anarchy, the destruction of our most cherished values—there is no need to do anything. We need only do nothing except what comes naturally—and breed. And how easy it is to do nothing. Isaac Asimov

The purpose of any training program is to provide a supply of skilled workers to address an unmet demand for their services. With respect to medical training, new physicians are required either to take the place of retiring physicians or to address an unmet demand for patient care. Evolving data strongly suggest that the supply of neurointerventional (NI) physicians is not only sufficient, but has exceeded the present need for services. Despite this, we continue to train new neurointerventionists (NIs) in unprecedented and increasing numbers every year. These new NI physicians are finding it progressively more difficult to secure employment and, once hired, face considerable challenges in building a practice and developing/maintaining their skills. Fellowship training is ingrained into the fabric of our academic practices and currently seems to be perpetuated more by inertia than a dynamic evaluation of the present workforce needs. It is the position of the authors that, if we do not re-evaluate this process, we are potentially doing a tremendous disservice to the people we are training, to patients in need of treatment (and maybe more importantly to those patients with lesions who are not in need of treatment), and finally to ourselves.

One of the more difficult aspects of evaluating the NI workforce is obtaining accurate data to characterize the status of current supply and demand—such as the number of practicing NI physicians, the number and growth rate of neuroendovascular cases and the number of new graduates entering the market each year. These statistics must be triangulated using several available sources, the most consistent of which seem to be data obtained by surveying vendors. For the purposes of our analysis of the demand for services, we will consider only aneurysm and acute stroke treatment since these represent the core endovascular procedures performed by almost all fellowship-trained NIs. Other NI procedures represent a relatively smaller proportion of the overall case volume (eg, tumor embolization, arteriovenous malformation embolization) or are performed by multiple non-neurologically-based specialties (eg, carotid stenting, spine interventions). We also did not consider diagnostic neuroangiography volume since interventional training is not required to perform this procedure.

How many practicing neurointerventionists are there in the USA right now? Answer: about 800

In 2008 there were 335 US hospitals at which six or more aneurysms were treated with endovascular coiling and 435 where two or more intra-arterial (IA) stroke treatments were performed.1 Assuming an average of two NI physicians per hospital, this yields between 670 and 870 NIs practicing in 2008. Clearly, a significant number of fellows have graduated and gone into practice since that time (∼50–80 per year), easily yielding around 800 or more practicing NIs currently in the USA.

Approximately 400 US hospital accounts are stocked with flexible distal access guiding catheters—an item essentially exclusively used for neurovascular intervention—again, yielding around 800 NIs provided that we assume there are two per account.

Number of aneurysms coiled each year? Answer: around 20 000–22 000

The Nationwide Inpatient Sample (NIS) provides data for the endovascular treatment of aneurysms through 2008. Approximately 20% of hospitals are included in the sample and the absolute case numbers can be multiplied by a factor of five to approximate the overall US case volume. The NIS data seem to significantly underestimate NI case volumes for the time periods covered. In addition, given that the most recent data are from 2008, the absolute case numbers may not be directly applicable to the present market (although trends within the NIS data are probably meaningful). Erring on the more optimistic side, we will use case volume estimates provided by vendors. Industry estimates for 2012 indicate that approximately 20 000–22 000 aneurysms will be treated with endovascular techniques this year in the USA. Using the number of physicians of 800 derived earlier, that yields about 26 aneurysms per year per operator, or one case every other week. Growth rates for endovascular aneurysm treatment have steadily declined as the penetration of coiling into clipping and the detection rate have all reached a steady state. As such, the current workforce of practicing NIs is more than sufficient to provide endovascular therapy for all of the cerebral aneurysms in the USA for the foreseeable future.

Number of ischemic stroke cases per year? Answer: approximately 11 000

Data from the NIS also provide what seem to be significant underestimates of the volume of interventional stroke cases currently being performed per year in the USA.2 ,3 Moreover, given the proliferation of thrombectomy devices over the past 5 years and the subsequent growth in case volume, the 2008 numbers probably have little applicability to the present situation. Again erring on the more generous side of case volume projections, we will use data from vendors. Industry sources estimate that between 10 000 and 12 000 IA stroke treatments (mostly thrombectomies) will be performed in 2012. For a population of 800 operators, this amounts to around 14 cases per operator per year, or less than one case every 3 weeks.

The mythical ever-approaching tsunami of IA stroke cases probably doesn't exist

The perception of some members of the medical community, which is reflected in the literature, is that the present volume of IA stroke interventions merely represents the ‘tip of the iceberg’ and that an explosive growth in case volume can be expected for years to come.4 However, a realistic look at the evolving clinical trial data, the present size and structure of established IA stroke programs and reasonable models for future growth provide a more sobering perspective on the potential ‘ceiling’ for interventional stroke case numbers.

We, the authors, all believe in the benefits of IA stroke intervention. However, we recognize that PROACT II is the only randomized clinical trial which supports IA stroke intervention and that these data do not apply to many of the patients who are currently receiving IA therapy.5 There are no randomized clinical trials which support the superiority of mechanical thrombectomy over medical therapy for the treatment of acute ischemic stroke caused by large vessel occlusion. This paucity of clinical evidence has resulted in some private insurers designating mechanical thrombectomy as ‘investigational and not medically necessary’, threatening the economic viability of these procedures by jeopardizing hospital/institutional reimbursement.6 The recent discontinuation of the Interventional Management of Stroke III study for futility only exacerbates this perception and will likely result in further scrutiny of IA stroke reimbursement by payers. The argument that additional NI physicians are needed is most often predicated upon the theoretical future growth of IA stroke therapy. This anticipated growth is completely dependent upon the favorable result of a randomized clinical trial that has yet to be completed.

Despite these considerations, robust ‘hub and spoke’ networks have been built to direct patients to interventional stroke centers, NIs are avidly lecturing local community healthcare practitioners about the availability of IA therapy at their respective centers and, in some cases, even providing ‘telestroke’ triage of patients in an attempt to facilitate the transfer of all potential candidates for IA therapy. These systems have been superimposed upon the pre-existing solid infrastructure built after the National Institute of Neurological Disorders and Stroke trial led to clearance by the Food and Drug Administration for the use of intravenous tissue plasminogen activator in acute ischemic stroke. Physicians and patients have been educated for almost two decades that stroke is an emergency and that efficiently delivered medical and/or interventional therapies can provide a dramatic potential for improvement. Once patients with large vessel occlusion, acute stroke symptoms and potentially salvageable brain reach the interventional centers, they are taken to catheterization laboratories for revascularization. As such, one must exercise caution when speculating that a positive IA stroke study would create a dramatic increase in the current interventional stroke case volume since, at most successful interventional centers, we are already functioning (in concert with our referring hospitals) as though a study proving the efficacy of IA thrombectomy has been completed! In other words, it is likely that the majority of patients to whom the results of a positive clinical study would pertain (eg, large vessel occlusion, high NIH Stroke Scale score, presenting <8 h after symptom onset, without a large completed stroke) are already primarily presenting to, or being transferred to, NI centers for possible IA thrombectomy.

Consider the argument made by Cloft et al in their analysis of interventional stroke. They propose that the needed number of interventional stroke centers should approximate the number of Level I trauma centers in the USA (around 200). It turns out that this is probably a generous estimate given that the lowest volume Level I trauma centers admit ∼1500 trauma cases per year while the ‘busiest’ stroke centers admit ∼1200 stroke patients per year and a ‘busy’ stroke center is characterized as having >400 stroke admits per year.7 So, if theoretically these 200 centers were to ramp up IA stroke volume to 100 cases per year, this would result in a total of only 20 000 cases per year.8 Hirsch et al arrive at a strikingly similar estimate using a different model. They hypothesize that the 500 busiest stroke centers in the USA will either be capable of delivering IA therapy or efficiently transferring their patients to a center where IA therapy is available. On the basis of their experience at the Massachusetts General Hospital, Hirsch et al calculated that around 5% of patients presenting primarily to their center as well as 5% of those transferred in from an outside institution were ultimately taken to the catheterization laboratory for IA treatment. If all patients presenting to the busiest 500 US stroke centers were provided this level of access to IA therapy and 5% of them were suitable candidates, this would result in approximately 10 400 IA stroke cases per year (essentially equal to the present volume of IA stroke cases). Even if the percentage of stroke patients taken for IA treatment were to double to 10%, the total number treated would only increase to ∼21 000 patients.9 This 100% increase from the present case volume would translate to an increase from one stroke case every 3.5 weeks to one every other week for each practicing NI.

One could argue that case volumes are not as important as the number of centers providing IA therapy since 24/7 coverage is required for each center. Provided that 200–300 interventional stroke centers could efficiently provide coverage for the entire USA and three physicians on average would be required to cover each of these institutions, a total of 600–900 NIs would be sufficient to provide IA stroke care to the entire US population. Currently, we have at least this number of practicing NIs and more than the needed number of interventional stroke centers to address these coverage needs.

Present workforce status: supply currently matches or exceeds existing and future demand for services

The present data provide a strong indication that the NI workforce is more than sufficient to cover the expected volume of cases for the foreseeable future. This assessment of an excess physician supply even allows for an optimistic projection which includes the successful completion of a trial confirming a benefit for interventional stroke treatment which would support the further build-out of the market to its full potential.

Yearly volume of graduating NI fellows? Answer: around 80–100

There are approximately 80 active training programs for NI fellows, each accepting on average between one and two fellows per year. As a point of comparison, there are only approximately 100 neurosurgical residency programs in the USA graduating a total of 160 residents each year. The volume of fellows in training has ballooned over the past decade as the numbers of fellowship programs have increased. Only a handful of these programs are accredited by the Accreditation Council for Graduate Medical Education (ACGME) and, as such, the vast majority are not held to any standard of diagnostic or therapeutic case volume. Thus, there is no assurance that these programs meet the ACGME standard of having ‘… the capacity … to offer an adequate educational experience in endovascular surgical neuroradiology …’. So, at this point, both the number and the quality of NI training positions are completely unregulated.

The field of NI surgery is new and many of its founding pioneers are still in active clinical practice today. In fact, the vast majority of NI operators have been in practice for less than 15 years. For this reason, the rate of retirement attrition in the NI specialty is exceedingly low.

At the present rate of training we are on pace to double the size of the NI workforce during this decade. If the number of active fellowship programs continues to grow at the current rate and in parallel with the NI physician workforce, these numbers will increase even more rapidly.

So what difference does it make that we have too many NI physicians?

The primary issue is that excess growth in the number of NIs invariably results in a decentralization of care. This dispersion of cases among a much larger number of operators and hospitals leads to a greater proportion of patients being treated by low-volume operators at smaller centers. There are numerous studies in the neurology, neurosurgery and NI literature which support the concept that patients treated at high-volume centers have significantly better outcomes than those treated at low-volume centers.10–13 Hoh et al demonstrated that patients with unruptured aneurysms undergoing endovascular treatment at high-volume institutions had fewer adverse outcomes, higher rates of discharge to home, shorter length of stay and lower total hospital charges. There was also a trend toward lower mortality rates.14 Gupta et al demonstrated that, for acute ischemic stroke treatment, large centers had shorter times from CT to groin puncture, shorter procedure times, higher rates of successful perfusion and better outcomes.15 So decanting case volumes from high-volume centers to low-volume centers has the potential to be detrimental to patient outcomes.

As the larger academic centers have become increasingly saturated with operators, graduating fellows accepting their first jobs are more frequently confronted with the daunting task of setting up a new program at a community hospital. An inexperienced NI physician working alone in the absence of supportive more senior partners in a smaller hospital with an inexperienced technologist and nursing staff and a very low case volume clearly does not represent an optimal environment in which to deliver patient care for complex cerebrovascular disease.

Moreover, as the numbers of proceduralists increase, there are increasing pressures—both economic and professional—upon new operators to find and aggressively treat patients in a highly competitive environment. As these pressures increase, it places considerable extrinsic bias upon decisions of treatment versus conservative management.16 ,17 That same pressure to generate and maintain adequate case volumes also acts as a direct disincentive for low-volume centers to transfer more complex cases to the regional centers of excellence who they view as direct competitors. As this environment evolves, it is difficult to deny that this might be detrimental to patient care and also disastrous for healthcare economics.

So what is the most responsible solution? Answer: suspend all NI fellowship training

Given the current data, it seems that there is no risk of ‘underserving’ the US population with NI physicians at any time in the near term. For this reason, we propose suspending the training of new fellows until the cerebrovascular community can be certain that a demand for additional NI physicans actually exists. We further submit that the need for additional NI physicians should be clearly defined on the basis of a reliable assessment of the following parameters:

  1. The existing workforce size: an accurate accounting of the number of practicing NIs of all specialties, the number of hospitals with NI programs and their geographic distribution throughout the USA.

  2. The existing demand for services: an accurate accounting of the yearly number of NI cases being performed in the USA.

  3. Yearly growth (or contraction) rates in overall NI procedure volumes.

While it might take some time to define these parameters to everyone's satisfaction, it would seem that time is certainly on our side here. In the absence of a randomized clinical trial supporting ‘state-of-the-art’ interventional stroke treatment, it seems implausible that case volumes will continue to grow and support any demand for additional NI physicians. On the contrary, a second futile or negative IA stroke study with modern devices could potentially end most interventional stroke therapy. For this reason, it is difficult to come up with a reason why additional physicians should be trained until we have a clearer view of the future (or at least a better handle on the present).

A second utility to this proposed ‘moratorium’ on training is that it will allow time for the societies to organize policies and procedures which mandate that future fellows will be trained only through pathways that are certified and governed by the ACGME. Thus, this suspension of training will allow the cerebrovascular community time to develop an accurate assessment of the workforce needs and to exercise greater control over the quality and number of fellowship positions offered in the future. Ultimately, this will help ensure that, going forward, we are meeting the existing demand with an appropriate supply of new well-trained physicians.

What are the potential downsides of halting training if we are incorrect in our calculations and the future need for services exceeds the current NI physician supply?

That scenario actually doesn't look all that bad. We would have to spend a larger percentage of our time doing NI cases; increased individual operator and hospital volumes would result in our becoming more proficient (which would be expected to improve patient outcomes); any perceived pressure to treat patients with marginal indications would dissipate; and relatively smaller centers with fewer operators would be more willing to transfer the most complex cases to larger centers. This ‘relative undersupply’ scenario seems quite favorable to patients, hospitals, healthcare economics and physicians, so it is hard to appreciate any risk.

A parting thought

Undeniably there are considerable infrastructural changes required for academic practices to transition away from fellowship training. However, the short-term challenges are well worth it if the long-term pay-off is the preservation of our specialty and the well-being of our patients. Fellowship training is a wonderfully rewarding and enjoyable experience, no doubt. In a way it's very similar to having children, including the consequences … just imagine if we all decided to have one or two of them per year … every year.


View Abstract


  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Linked Articles

  • PostScript
    Buddy Connors
  • PostScript
    David A Rubin
  • PostScript
    Lee-Anne Slater Winston Chong Michael Holt Ronil V Chandra