PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Arch Intern Med. Author manuscript; available in PMC 2012 November 14.
Published in final edited form as:
PMCID: PMC3213299
NIHMSID: NIHMS297832

Physician Specialty and Carotid Stenting Among Elderly Medicare Beneficiaries in the United States

Abstract

Background

The use of carotid stenting is rising across the United States. How physician specialty relates to its utilization or outcomes is uncertain.

Methods

We performed an observational analysis of fee-for-service Medicare beneficiaries 65 years or older undergoing carotid stenting between 2005 and 2007 in 306 hospital referral regions (HRRs). We first determined how frequently carotid stenting was performed by different specialists within each HRR, and then used multivariable regression models to compare population-based utilization rates and 30-day outcomes for it across HRRs based on the proportion performed by cardiologists, surgeons, radiologists or a mix of specialists.

Results

In 272 HRRs where at least 15 cases were performed during the study period, we identified 28,700 carotid stenting procedures performed by 2588 operators. While cardiologists made up approximately one-third of these operators, they were responsible for 14,919 (52.0%) procedures. Significant differences were noted in the characteristics of patients treated by cardiologists as compared with other specialties, including higher rates of invasive cardiac procedures and lower rates of acute stroke or transient ischemic attacks in the 180 days prior to carotid stenting. Population-based utilization rates were significantly higher in HRRs where cardiologists performed most procedures relative to HRRs where most were done by other specialists or a mix of specialists (P<0.001). In contrast, risk-standardized outcomes did not appear to differ across HRRs based on physician specialty.

Conclusion

Carotid stenting is being performed by operators from diverse specialties. HRRs where cardiologists perform most procedures have higher population-based utilization rates with similar outcomes.

Nearly 100,000 Americans undergo carotid endarterectomy each year with the goal of reducing their risk of stroke.1 This operative procedure is performed by surgeons, traditionally following diagnostic carotid and cerebral angiography by radiologists. A promising and innovative alternative to carotid endarterectomy is carotid stenting, which is a minimally-invasive, endovascular procedure that may be of particular benefit in high-risk patients.2 Since its initial approval in late 2004, use of carotid stenting has grown rapidly across the United States3,4 with enthusiastic adoption by operators from several, diverse physician specialties including surgery, radiology and cardiology.5

While its performance by surgeons and radiologists is not unexpected, the large numbers of cardiologists involved with carotid stenting has been somewhat more surprising – especially given their limited prior experiences with carotid disease or stroke care in general. Yet for many reasons a cardiologist performing carotid stenting makes perfect sense.6,7 Cardiologists care for many elderly patients with atherosclerotic heart disease so their involvement could lead to greater recognition and earlier treatment of carotid disease. Extensive experience with catheter-based techniques in the coronary arteries also makes it potentially straightforward for them to learn the skills required for it. Yet the involvement of cardiologists in carotid stenting is not without concern. The overall benefits of carotid stenting remain controversial,8 and a growing role of cardiologists in other non-cardiac vascular interventions has led to increases in their overall utilization with no differences in patient outcomes.9,10

Accordingly, we examined national data on elderly Medicare beneficiaries to determine: (1) the current specialty composition of operators performing carotid stenting in the United States; and (2) the extent to which specialty explains variation in its utilization and outcomes across different healthcare markets. For the reasons mentioned above, we focused on comparing carotid stenting performed by cardiologists with those done by other specialists. Understanding how physician specialty may affect the utilization of a novel procedure like carotid stenting has important implications for optimizing its dissemination.

Methods

Data Sources and Study Cohort

From the Centers for Medicare & Medicaid Services (CMS), we obtained Physician Carrier (Part B), Medicare Provider Analysis and Review (MEDPAR) and Denominator files from January 1, 2005 to December 31, 2007. Physician Carrier files include data on claims for procedures from non-institutional providers, such as physicians, while MEDPAR files include data on acute-care hospitalizations and Denominator files contain information on eligibility and enrollment. Data on all Medicare beneficiaries aged 65 years or older enrolled in fee-for-service programs within the United States were included. The Institutional Review Board of the University of Michigan approved this protocol prior to its initiation.

From the Physician Carrier files, we identified all procedures of carotid stenting using Healthcare Common Procedure Coding Systems (HCPCS) codes 37215 and 37216. We determined the specialty of the operator performing carotid stenting using the Medicare specialty code associated with the procedure and then organized them into 3 categories: 1) those performed by cardiologists, 2) those performed by surgeons, and 3) those performed by radiologists. Surgeons included vascular and general surgeons as well as other surgical specialties (e.g., cardiovascular surgeons, neurosurgeons). In cases when a specialty code was unreported or reflected broader fields of practice (i.e., general practice or internal medicine), we searched the 2007 Unique Physician Identification Number (UPIN) Directory to determine if a more narrow specialty (e.g., cardiology) was associated with that operator. Although uncommon, procedures performed by operators from specialties other than the 3 listed above (n=679) or those performed by 2 or more operators from different specialties (n=150) were included in a separate category.

Information on age, gender, and race were also obtained for each procedure from the Physician Carrier files. For the subset of patients continuously enrolled in fee-for-service programs at least 1 year prior to their procedure (n=26,889), we reviewed all claims in the Physician Carrier and MEDPAR files to collect data on the presence of comorbidities. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnostic codes associated with these claims were used to calculate an Elixhauser co-morbidity score. We also used these claims to determine if in the 180 days prior to carotid stenting a patient had: (1) undergone a cardiac catheterization (with or without concomitant carotid x-ray angiography) or coronary intervention; or (2) recent diagnosis of acute stroke or transient ischemic attack (TIA) based on a hospitalization or physician claim. A 180 day time period was chosen given that this length of time was also used by the Carotid Revascularization Endarterectomy vs. Stenting Trial (CREST) investigators to define symptomatic disease.11

Hospital Referral Regions and Rates of Population-Based Utilization and Outcomes

Each procedure of carotid stenting was mapped to a hospital referral region (HRRs) based on the patient's residence Zip code using the Dartmouth Atlas of Cardiovascular Health Care.12 HRRs are large geographic units representing distinct healthcare markets for tertiary care. After assigning each procedure to 1 of 306 HRRs and determining the specialty of the operator that performed it, we further categorized HRRs into the following 3 groups: (1) those where cardiologists performed most (or >50%) carotid stenting; (2) those where surgeons performed most carotid stenting; (3) those where radiologists performed most carotid stenting. “Mixed” HRRs where no specialty performed most carotid stenting were included as a separate group. Only data from HRRs where 15 or more procedures were performed during the study period were included.

We calculated population-based utilization rates for carotid stenting in each HRR over the 3-year study period. Numerators for these rates were the total number of procedures performed on patients within the HRR while denominators were the total number of beneficiaries within the HRR at the midpoint of each year within the study period. Utilization rates were then adjusted for differences in age (65–69, 70–74, and 75 years or older), sex, and race (black, nonblack) across HRRs using direct standardization techniques and subsequently averaged over the 3-year period. For each HRR, we also calculated: (1) 30-day risk-standardized mortality and (2) 30-day risk-standardized mortality or stroke admission. We determined death at 30-days using Denominator files and stroke admission using the MEDPAR files. Risk-standardization was performed using 2-level models that accounted for clustering at the HRR-level and adjusted for: age, gender, race, recent diagnosis of acute stroke or TIA, Elixhauser co-morbidity score, embolic protection device use, and total volume of the operator performing the procedure over the study period. From these models, we calculated predicted and expected 30-day rates for each HRR for both outcomes in order to generate a standardized ratio. We multiplied this ratio by the overall 30-day mortality rate and the 30-day mortality or stroke admission rate in the study population to generate risk-standardized rates.

Statistical Analysis

Analyses were performed at both the procedural and HRR-level. Descriptive statistics evaluated for differences in the characteristics of patients undergoing carotid stenting across physician specialties. We then examined how adjusted utilization rates and 30-day risk-standardized outcomes for carotid stenting varied across HRRs where cardiologists performed <25%, 25 to 50%, and >50% of the procedures. Multivariable linear regression models were then used to assess if rates of utilization and risk-standardized 30-day mortality differed between HRRs where most procedures were performed by cardiologists relative to HRRs where most procedures were performed by surgeons or radiologists, or Mixed HRRs. At the HRR-level, these models accounted for: (1) the number of operators performing carotid stenting per 100,000 enrollees; (2) the geographic region (Northeast, South, Midwest, West); and (3) the age, gender, and race-adjusted rate of carotid endarterectomy in 2004. Although imperfect, we attempted to account for the overall prevalence of carotid disease in the HRR as well as overall “enthusiasm” for referral for carotid revascularization in that region by using adjusted utilization rates of carotid endarterectomy in the HRR during 2004 – the year prior to the widespread introduction of carotid stenting.

All analyses were performed using SAS version 9.2 (SAS Institute, North Carolina) and Stata version 11.0 (StataCorp, Texas). P values of <0.05 were considered significant and all tests were 2-sided.

Results

In 272 HRRs where at least 15 cases were performed, we identified 28,700 carotid stenting procedures performed in 26,938 patients between January 1, 2005 and December 31, 2007. Of these, 14,919 (52.0%) procedures were performed by cardiologists, 7840 (27.3%) by surgeons and 5112 (17.8%) by radiologists, with the remaining 829 (3.2%) performed by other specialties that were largely represented by neurology and internal medicine. Of the operators performing carotid stenting, 904 (34.9%) were cardiologists, 864 (33.4%) were surgeons, 719 (27.8%) were radiologists and 101 (3.9%) were other specialists. Among surgeons, vascular surgeons made up the largest group (n=539 [62.4%]) followed by general surgeons (n=180 [20.8%]), neurosurgeons (n=68 [7.9%]) and cardiothoracic surgeons (n=68 [7.9%]). The number of procedures performed during the study period grew across all specialties but was most prominent for cardiologists and surgeons (Figure 1).

Figure 1
Cumulative number of procedures of carotid stenting over time by specialty.

Table 1 displays the characteristics of patients undergoing carotid stenting across specialties. No significant differences were seen by gender or race across procedures performed by cardiologists relative to other specialists, although modest differences in age were noted across groups. The overall mean Elixhauser co-morbidity scores were slightly higher in procedures performed by cardiologists but there were key differences in the presence of specific conditions that made up the Elixhauser co-morbiditiy scores. For example, cardiologists were more likely to perform procedures in patients with Elixhauser co-morbidities related to cardiovascular conditions but less likely to do so in those with neurological conditions. Finally, rates of cardiac catheterization (including with concomitant carotid x-ray angiography) and coronary intervention in the 180 days prior to carotid stenting were significantly higher in patients undergoing carotid stenting by cardiologists. However, rates of recent diagnosis of acute stroke or TIA in the 180 days prior to carotid stenting were much less in this group.

Table 1
Patient characteristics for carotid stenting across physician specialties.

The proportion of carotid stenting performed by cardiologists within each HRR varied substantially across markets, with a mean of 45.5% that ranged from 0% in 2 HRRs to 100% in 1 HRR. Cardiologists performed most procedures in 115 (42.3%) of the 272 HRRs included in this analysis (Figure 2). By comparison, surgeons performed most procedures in 41 (15.1%) HRRs and radiologists in 37 (13.6%) HRRs. The remaining 79 Mixed HRRs had no single specialty performing most procedures. Characteristics of the HRRs are displayed in Table 2 stratified by the specialty that performed most carotid stenting in that market. Overall, HRRs where cardiologists performed most procedures were more frequently located in the Midwest and South, and had a higher number of operators performing carotid stenting per 100,000 enrollees.

Figure 2
Map of HRRs where cardiologists, surgeons and radiologists perform most carotid stenting.
Table 2
Characteristics of HRRs categorized by the physician specialty that performed most carotid stenting in that market.

The overall age, gender and race-adjusted rate of carotid stenting was 3.3 (± 2.4) per 10,000 enrollees during the study period. Substantial variation existed in its use across HRRs with adjusted utilization rates varying from 0.3 to 20.2 per 10,000 enrollees. HRRs where cardiologists performed most carotid stenting had significantly higher adjusted utilization rates than those where cardiologists performed 25% or fewer (4.3 per 10,000 enrollees versus 2.5 per 10,000 enrollees; P<0.001) (Figure 3), but no difference in 30-day risk-standardized mortality (1.8% versus 1.8%; P=0.12) or 30-day risk-standardized mortality or stroke admission (4.5% versus 4.7%; P=0.16). After accounting for the number of operators, geographic region, and baseline rates of carotid endarterectomy in 2004, HRRs where cardiologists performed most carotid stenting had significantly higher adjusted utilization rates relative to HRRs where most were done by other specialists or a mix of specialists but no differences in risk-standardized outcomes (Table 3).

Figure 3
Mean rate of carotid stenting per 10,000 enrollees in hospital referral regions (HRRs) based on the proportion performed by cardiologists. HRRs where cardiologists performed most carotid stenting had higher adjusted utilization rates than others(P<0.001). ...
Table 3
Characteristics of HRRs categorized by the physician specialty that performed most carotid stenting in that market.

Comment

Our study has 3 noteworthy findings. First, we found cardiologists in the United States currently play a substantial role in carotid stenting. Approximately one-third of the operators performing this procedure in Medicare beneficiaries are cardiologists and they are responsible for over half the total procedures. Surgeons and radiologists, in contrast, were responsible for 28% and 18% of procedures, respectively. Second, we found that cardiologists treat patients substantially different from other specialists. Not surprisingly, procedures by cardiologists more frequently involved patients with cardiac conditions or recent invasive cardiac procedures. Yet these patients also had fewer neurologic conditions, including less evidence of recent acute stroke or TIA. While it is not possible to definitively assess for symptomatic disease in these data, the differences we found across physician specialty were substantial. This finding will be important to clarify in future work since the benefits of carotid revascularization are generally less in asymptomatic patients. Finally, we found that HRRs where cardiologists performed most carotid stenting had higher population-based rates of utilization than other HRRs with similar outcomes.

Unlike many medical procedures, carotid stenting is a “cross-disciplinary” technology that is performed by operators from several, diverse physician specialties. These groups have widely different clinical opportunities and incentives for performing it, which could lead to important differences in how patients are selected for carotid stenting. For example, cardiologists care for a large number of patients with atherosclerotic heart disease. These individuals may also have coexisting carotid disease that may or may not be symptomatic. Involvement of cardiologists in the care of these patients could lead to greater recognition of carotid stenosis in the general population and subsequent referral for carotid stenting, particularly among less symptomatic patients. For example, we found that over 50% of patients who were treated by cardiologists and also underwent cardiac catheterization prior to carotid stenting had carotid and cerebral angiography performed simultaneously. This proportion was much lower among those treated by surgeons (24%), radiologists (18%) or others (34%), which suggests the possibility that routine case finding of severe carotid stenosis by cardiologists during diagnostic angiography influenced patient selection (i.e., “oculo-stenotic reflex”).

Radiologists, in contrast, typically have little involvement in the early evaluation of patients with stroke and carotid stenosis and rely primarily on referrals from other physicians. Surgeons are the only operators involved in carotid stenting that also perform carotid endarterectomy – the open-surgical procedure that is the current standard for carotid revascularization. This may make them less likely to view carotid stenting as a first-line therapy and restrict their use to select cases. Differences in educational training may also contribute to their overall decision to perform carotid stenting. For example, cardiologists routinely use standard 0.014” wire-based equipment and embolic protection devices for their coronary interventions, both of which are requirements for most carotid stenting. It is possible that these experiences make them generally more enthusiastic for the use of catheter-based techniques than surgeons or even radiologists. Prior work also supports this possibility as the involvement of cardiologists in other non-cardiac vascular interventions has been linked to higher population-based rates of utilization for those procedures.9,10

Understanding the role of physician specialty in carotid stenting is not just of theoretical interest, but importantly, has practical implications. In 2006, the Centers for Medicare & Medicaid Services (CMS) raised considerable controversy when it proposed modifying its current national coverage decision for carotid stenting to allow only physicians who perform carotid endarterectomy to determine which patients would be suitable for carotid stenting.13 Although the goal was to improve overall decision-making for carotid stenting, cardiologists and radiologists were concerned that this would have essentially left surgeons as the lone gatekeepers for its use. Although CMS subsequently retreated from this proposed modification, our findings suggest that decisions about using this procedure may indeed be linked to physician specialty. If this is true, promoting multi-disciplinary decision-making around carotid stenting could reduce unwanted variation in its adoption. The inclusion of specialists not directly involved in carotid stenting, like neurologists, may also aid in these decisions, particularly since they are more experienced in identifying stroke-related complications and managing these patients long-term.14

These issues all take on even greater relevance given current controversies surrounding the clinical effectiveness of carotid stenting relative to carotid endarterectomy. For example, the CREST investigators recently reported that both procedures led to similar rates of stroke, myocardial infarction or death in 2502 patients over a mean follow-up period of 2.5 years.11 Although this appears to suggest overall equivalence between the 2 strategies, the large International Carotid Stenting Study (ICSS) – also published earlier this year – found higher rates of this combined endpoint with carotid stenting.15 In addition, both trials found a significant interaction between age and outcomes that favored carotid endarterectomy in older patients, which is of direct relevance for the elderly Medicare population we studied. With this degree of uncertainty about the role of carotid stenting, it is likely that substantial variation in its use will continue and could intensify in regions more enthusiastic for its application.16

Our study should be interpreted with the following limitations in mind. We relied on Medicare claims data, which are inherently limited in their ability to understand the clinical context in which a procedure may be used. For example, we could not directly determine whether patients were “symptomatic” or “asymptomatic” and had to rely on the presence of claims data suggesting a recent diagnosis of acute stroke or TIA. As a result, we are unable to comment on the overall appropriateness of its use across specialties despite wide variation in its adoption. It may be that markets where cardiologists are performing most procedures are “overutilizing” carotid stenting or that markets where surgeons or radiologists predominately perform the procedure have important barriers that limit access and are “underutilizing” it. Importantly, either concern may be overcome by promoting multidisciplinary decision-making to optimize its utilization. Along similar lines, it could be argued that differences in adoption of carotid stenting across markets may be due to variation in the prevalence of carotid disease or the overall enthusiasm of physicians for carotid revascularization within in a region. Although admittedly an imperfect measure, we attempted to account for these factors in our analysis by using adjusted utilization rates for carotid endarterectomy in 2004. A stronger counterargument to this concern, however, are recent data reporting large differences in rates of carotid stenting across markets – a degree of variation that is unlikely to be explained by differences in these factors alone.3,4

Another consideration is that findings in Medicare claims data may not be generalizable to younger patients or managed care populations. Although some of these concerns could be overcome using data from clinical registries, those sources have their own limitations. They are typically limited to select high-volume centers and have no information on “at-risk” denominator populations to determine how its use by different specialties might impact on utilization. An additional complicating factor is that different professional organizations (e.g., the American College of Cardiology and the Society for Vascular Surgery) are promoting unique registries, often to their own physician constituents.17,18 This may prevent these registries from providing a comprehensive picture of carotid stenting in the United States and limits their ability to address policy-based questions across different specialties and broad patient populations. In contrast, this is a particular advantage of using Medicare claims data. Finally, the most reliable outcome we were able to evaluate was mortality, which is generally low following carotid stenting. Our ability to assess other important outcomes, like the development of stroke, is more limited within claims data.

In conclusion, cardiologists perform the majority of carotid stenting in the United States. Procedures performed by cardiologists involve patients who appear to differ in important ways from those treated by other specialists like surgeons and radiologists. HRRs where cardiologists perform most procedures have significantly higher population-based rates of utilization suggesting an important role for physician specialty in the choice to use carotid stenting. Further studies are needed to better understand the specific nature of this relationship and whether multi-disciplinary decision-making by teams of specialists could optimize the use of this innovative technology.

Acknowledgment

We are grateful to Onur Baser, PhD and Li Wang, PhD, for their roles in preparing data. Neither of these individuals received compensation for this work outside of their usual salaries. The authors have no financial conflicts of interest to disclose in regards to this work. This project was supported by a grant from the National Institutes of Health (5R21AG032155-02). The funding agency was not involved in the design and conduct of the study; in data management or analysis; or in manuscript preparation. The views expressed in this article are those of the authors and do not necessarily represent the views of this agency or the Department of Veterans Affairs and the Department of Health and Human Services. Dr. Nallamothu had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

References

1. Lloyd-Jones D, Adams RJ, Brown TM, et al. Heart Disease and Stroke Statistics--2010 Update: A Report From the American Heart Association. Circulation. 2010;121(7):e46–215. [PubMed]
2. Bates ER, Babb JD, Casey DE, et al. ACCF/SCAI/SVMB/SIR/ASITN 2007 clinical expert consensus document on carotid stenting: a report of the American College of Cardiology Foundation Task Force on Clinical Expert Consensus Documents (ACCF/SCAI/SVMB/SIR/ASITN Clinical Expert Consensus Document Committee on Carotid Stenting) J. Am. Coll. Cardiol. 2007;49(1):126–170. [PubMed]
3. Patel MR, Greiner MA, DiMartino LD, et al. Geographic Variation in Carotid Revascularization Among Medicare Beneficiaries, 2003–2006. Arch Intern Med. 2010;170(14):1218–1225. [PubMed]
4. Goodney PP, Travis LL, Malenka D, et al. Regional Variation in Carotid Artery Stenting and Endarterectomy in the Medicare Population. Circ Cardiovasc Qual Outcomes. 2010;3(1):15–24. [PubMed]
5. Rosenfield K, Babb JD, Cates CU, et al. Clinical competence statement on carotid stenting: Training and credentialing for carotid stenting--multispecialty consensus recommendations: A report of the SCAI/SVMB/SVS Writing Committee to develop a clinical competence statement on carotid interventions. J Am Coll Cardiol. 2005;45(1):165–174. [PubMed]
6. Gray WA. A cardiologist in the carotids. J Am Coll Cardiol. 2004;43(9):1602–1605. [PubMed]
7. Wholey MH, Wholey M, Mathias K, et al. Global experience in cervical carotid artery stent placement. Catheter Cardiovasc Interv. 2000;50(2):160–167. [PubMed]
8. Davis SM, Donnan GA. Carotid-artery stenting in stroke prevention. N Engl J Med. 2010;363(1):80–82. [PubMed]
9. Axelrod DA, Fendrick AM, Birkmeyer JD, Wennberg DE, Siewers AE. Cardiologists performing peripheral angioplasties: impact on utilization. Eff Clin Pract. 2001;4(5):191–198. [PubMed]
10. Murphy TP, Soares G, Kim M. Increase in utilization of percutaneous renal artery interventions by medicare beneficiaries, 1996–2000. AJR Am J Roentgenol. 2004;183(3):561–568. [PubMed]
11. Brott TG, Hobson RW, Howard G, et al. Stenting versus endarterectomy for treatment of carotid-artery stenosis. N Engl J Med. 2010;363(1):11–23. [PMC free article] [PubMed]
12. Wennberg DE, Birkmeyer JD, editors. The Dartmouth Atlas of Cardiovascular Health Care. American Hospital Association Press; Chicago, Ill: 1999.
13. [[Accessed August 28, 2010]];Decision Memo for Percutaneous Transluminal Angioplasty (PTA) of the Carotid ARtery Concurrent with STenting (CAG-00085R3) Available at: http://www.cms.gov/mcd/viewdecisionmemo.asp?from2=viewdecisionmemo.asp&id=194&. [PubMed]
14. Zaidat OO, Qureshi AI. Decision Memo for Percutaneous Transluminal Angioplasty of the Carotid Artery Concurrent with Stenting (CAG-00085R3) J Neuroimaging. 2008;18(3):237–238. [PubMed]
15. Ederle J, Dobson J, Featherstone RL, et al. Carotid artery stenting compared with endarterectomy in patients with symptomatic carotid stenosis (International Carotid Stenting Study): an interim analysis of a randomised controlled trial. Lancet. 2010;375(9719):985–997. [PMC free article] [PubMed]
16. Halm EA. The Good, the Bad, and the About-to-Get Ugly: National Trends in Carotid Revascularization. Arch. Intern. Med. 2010;170(14):1225–1227. [PubMed]
17. Sidawy AN, Zwolak RM, White RA, et al. Risk-adjusted 30-day outcomes of carotid stenting and endarterectomy: results from the SVS Vascular Registry. J. Vasc. Surg. 2009;49(1):71–79. [PubMed]
18. Anderson HV, Rosenfield KA, White CJ, et al. Clinical features and outcomes of carotid artery stenting by clinical expert consensus criteria: a report from the CARE registry. Catheter Cardiovasc Interv. 2010;75(4):519–525. [PubMed]