Restenosis

Jump to navigation Jump to search

Restenosis Microchapters

Home

Patient Info

Overview

Historical Perspective

Pathophysiology

Epidemiology & Demographics

Risk Factors

Screening

Causes of Restenosis

Differentiating Restenosis from other Diseases

Natural History, Complications & Prognosis

Diagnosis

History & Symptoms

Physical Examination

Lab Tests

Electrocardiogram

Chest X Ray

CT

MRI

Echocardiography or Ultrasound

Other Imaging Findings

Other Diagnostic Studies

Treatment

Medical Therapy

Surgery

Primary Prevention

Secondary Prevention

Cost-Effectiveness of Therapy

Future or Investigational Therapies

Restenosis On the Web

Most recent articles

Most cited articles

Review articles

CME Programs

Powerpoint slides

Images

American Roentgen Ray Society Images of Restenosis

All Images
X-rays
Echo & Ultrasound
CT Images
MRI

Ongoing Trials at Clinical Trials.gov

US National Guidelines Clearinghouse

NICE Guidance

FDA on Restenosis

CDC on Restenosis

Restenosis in the news

Blogs on Restenosis

Directions to Hospitals Treating Restenosis

Risk calculators and risk factors for Restenosis

Editor-In-Chief: C. Michael Gibson, M.S., M.D. [1], Associate Editor(s)-In-Chief:: Bhaskar Purushottam, M.D. [2]

Overview

Restenosis literally means the reoccurrence of stenosis. This is usually restenosis of an artery, or other blood vessel, but possibly any hollow organ that has been "unblocked". This term is common in vascular surgery, cardiac surgery, interventional radiology, or interventional cardiology following angioplasty, all branches of medicine that frequently treat stenotic lesions. In simple words, coronary re-stenosis can be considered as the reduction in the lumen diameter after a percutaneous coronary intervention (PCI), which induces iatrogenic arterial injury and results in neointimal tissue proliferation.[1] It can be defined based on angiography or as clinical restenosis. By angiography, the term 'Binary Angiographic Re-stenosis' is defined as > 50% luminal narrowing at follow-up angiography.[2] However, the most widely accepted and relevant definition would be a 'Clinical Re-stenosis', which is defined as need for a repeat target lesion revascularization (TLR) due to symptomatic coronary ischemia from the previously intervened vessel (proposed by the Academic Research Consortium). Therefore, this definition needs angiographic narrowing as well as clinical correlation. If the lesion does not meet angiographic criteria, but meets the criteria for a physiologically significant lesion by fractional flow reserve (FFR) or anatomically by intravascular ultrasound (IVUS) with the appropriate clinical context, it is still considered 'Clinical Re-stenosis'. PCI has evolved significantly from plain balloon angioplasty to the development of biodegradable stents in the last few decades. Currently, almost all coronary interventions use a bare metal stent (BMS) or more so a drug eluting stent (DES). Hence, the discussion in the following paragraphs will focus on in-stent re-stenosis of drug eluting and bare metal stents.

Coronary Restenosis

There are probably several mechanisms that lead to restenosis. An important one is the inflammatory response, which induces tissue proliferation around an angioplasty site.

Cardiologists have tried a number of approaches to decrease the risk of restenosis. Stenting is becoming more commonplace; replacing balloon angioplasty. During the stenting procedure, a metal mesh (stent) is deployed against the wall of the artery revascularizing the artery. Other approaches include local radiotherapy and the use of immunosuppressive drugs, coated onto the stenting mesh. Analogues of rapamycin, such as tacrolimus (FK-506), sirolimus and more so everolimus, normally used as immunosuppressants but recently discovered to also inhibit the proliferation of vascular smooth muscle cells, have appeared to be quite effective in preventing restenosis in clinical trials. Antisense knockdown of c-myc, a protein critical for progression of cell replication, is another approach to inhibit cell proliferation in the artery wall and has been through preliminary clinical trials using Morpholino oligos.

Histopathology

Restenosis can considered a local vascular manifestation of the general biological response to arterial wall injury. Following vessel wall injury with endothelium damage, the exposure of tunica media to the flowing blood results in platelet adhesion to the media's collagen via the Von Williebrand factor. This results in platelet activation, which is followed by aggregation and soon evolving into a vicious cycle, which feeds and nurtures itself till a clot is formed from the coagulation cascade (which is activated from the tissue thromboplastin released from the media following vessel wall injury) and fibrin. Basically, this phenomenon can be considered synonymous to hemostasis. Activated platelets express adhesion molecules, to which circulating leukocytes attach and begin the process of leukocyte migration. Also during this process, several chemotactic factors are released and multiple adhesion molecules are expressed by various activated inflammatory cells in the vicinity of the inflammation, which cause chemotaxis of leukocytes to the core of the inflammation. Growth factors, cytokines and adhesion molecules released and expressed by platelets, leukocytes and other inflammatory cells, result in the migration of vascular smooth muscle cells from the media and proliferation with resultant neo-intima formation [3] [4] [5]. Hence, neo-intima consists of vascular smooth muscle cells, macrophages and extracellular matrix, which has formed over several weeks.

When reading through the above process, we realize that plaque rupture and stent thrombosis uses the same tools of inflammation mentioned above. The phenomenon of stent thrombosis and acute coronary syndrome can be thought of as an acute phenomenon, evolving in minutes to a couple of hours. Thus, depending on several factors (which favor acute thrombosis) the above mentioned process can evolve very quickly resulting in thrombosis of the coronary vessel. This process can then follow any of the pathways: 1) the patient seeks immediate help and the percutaneous coronary intervention can open up the vessel and prevent progression of the acute coronary syndrome 2) the patient fails to seek help and can die from a massive myocardial infarction 3) sometimes, without seeking help, the patient lives through this coronary thrombosis and the clot becomes organised and recanalised. In more simple words, coronary restenosis can be thought of as a sub-acute to chronic vascular inflammatory response to iatrogenic controlled injury; while stent thrombosis is an acute vascular response resulting in acute thrombosis and a high propensity to cause sudden death or significant morbidity.

In balloon angioplasty, the lumen size is increased from its baseline diseased state through the dilatation of the balloon at the site of the diseased region. However, the elastic recoil (of the artery), negative remodeling, contraction and finally neo-intima formation, results in restenosis. Also, to be noted is the fact that there is a 30% risk of abrupt closure of the vessel from acute thrombosis from plain balloon angioplasty as it is a form of vascular injury too (iatrogenic controlled injury). Based on individual clinical and angiographic characteristics, some patients develop this fatal complication, while others do not.

In stent implantation, the lumen size is increased from its baseline diseased state and has a better patency than plain balloon angioplasty, as the stent scaffold holds the vessel open; therefore the component of elastic recoil and negative remodeling is reduced significantly. Thus the restenosis rates are lower in BMS when compared to plain balloon angioplasty, as the elastic recoil and negative remodeling is virtually eliminated. Hence, the final lumen size is greater and neo-intima formation does not materialize into clinical restenosis as seen to occur more commonly with balloon angioplasty. Similarly, it is well known that DES has a lower re-stenosis rate than BMS as the DES have the added advantage of releasing anti-stenotic drug which can retard the neo-intima formation. Once again, there are several other factors involved in re-stenosis of BMS and DES (as discussed below). The discussion of stent thrombosis in DES and BMS is beyond the scope of this chapter.

Coronary re-stenosis can be considered in two different settings of a percutaneous coronary intervention:

1. In pure balloon angioplasty (in the absence of stenting) Following balloon angioplasty, there is injury to the vessel wall, which triggers the expected inflammatory response from the vessel wall. Hence, there is elastic recoil, negative remodelling (reduction in lumen size due to the healing process) or contraction, thrombus at the site of injury (i.e., release of prothrombotic thromboplastin like material from the vessel wall, exposure of the collagen of the media), smooth muscle proliferation and migration and excessive extracellular matrix production. Neointima is composed principally smooth muscle cells and extracellular matrix.

2. With stenting Stenting prevents elastic recoil and negative remodelling. However, neointimal formation progresses and causes re-stenosis, which is reduced with antistenotic drugs. Restenosis in patients after stent implantation is caused predominantly by neointimal hyperplasia. However neointima formation in plain balloon angioplasty has a minor role.[6]

The porcine model of ISR has demonstrated the formation of a thick neointima in 28 days-which may be the reason as to why the earlier DES released their antistenotic drug by the end of 30 days. However, it is well known that the peak period of ISR development in humans is approximately 3 to 6 months-hence the newer DES are aiming to release the drug for an extended period. Interestingly, in these porcine models they found a positive correlation between the inflammatory infiltrate, degree of arterial injury and extent of stent strut penetration into the vessel wall with the neointimal thickness. Interestingly, in a rat model of carotid artery dilatation the gate-keeper step in initiating vascular smooth muscle cell proliferation from tunica media to intima is the rupture of internal elastic lamina.[7]

The above mentioned responses and time periods were in porcine models with near normal coronaries and this may not be the case in humans who have had balloon angioplasty, BMS and DES for atherosclerotic lesions. The inflammatory response in DES is very different in terms of cell composition and timing, when compared to balloon angioplasty and BMS and also among the different DES. These differences could be explained by the presence of various forms of polymer among DES. For example, the inflammatory response (predominantly giant cell infiltrates) to Sirolimus DES has been shown to persist beyond 180 days and up to 2 years. In contrast, the inflammatory response to BMS and the second generation Everolimus DES (which has a more biocompatible polymer) has been limited to 90 days and 12 months, respectively. Evidence of such persistence of inflammatory response has been found in autopsy cases and from thrombus aspirates taken from patients during percutaneous coronary intervention for late stent thrombosis.

Pathophysiology

DES predominantly consist of 3 elements: 1. Scaffold or the stent platform, which forms the skeleton of the stent 2. The antistenotic drug (such as paclitaxel, everolimus, etc.) 3. The polymer or the carrier on which the drug is mounted.

Unlike, the DES the BMS consist only the scaffold or the platform.

The pathophysiology seems to be the interplay of clinical characteristics, biological factors, mechanical factors, technical factors and finally de-novo lesions which arise within the stent itself.

The various pathophysiological factors are discussed as follows:

A) Biological and Genetic Factors

1. Drug Resistance Recent studies have revealed a genetic basis for drug resistance. Depending on the penetrance and expressivity of these mutations, the sensitivity to these drugs vary. This resistance can be the result of inherited genetic mutations or acquired following the exposure of a cytotoxic drug.

2. Genetic Factors Affecting the Inflammatory Response Polymorphism of glycoprotein IIIa and a mutant form of methylenetetrahydrofolate reductase appear to increase the risk of ISR. Interestingly, allele 2 of interleukin IL-1ra gene appears to be protective. However, these are just some of the genetic factors which have been identified in the causation of ISR. Hence, there remains the possibility of complex multigenic abnormalities (interplay of multiple variant forms of genes and environmental factors), which can play a significant role in ISR. These identified and yet to be identified genetic factors may explain as to why some patients develop ISR and some don’t despite identical clinical factors and stent characteristics.

3. Hypersensitivity The implantation of these stents is recognized as foreign antigens and hence a hypersensitivity reaction can be triggered, which can lead to ISR. As mentioned earlier, the DES has 3 components and the BMS has only the scaffold, which can all contribute towards this hypersensitivity reaction.

a. Stent Platform The stent platforms of the BMS and first generation DES (paclitaxel and sirolimus) is made up of 316L stainless steel, which contains more nickel and molybdenum than the second generation DES, where the scaffold is made up of cobalt chromium. This nickel and molybdenum are known to trigger these hypersensitivity reactions. Interestingly, to date, no prospective studies have confirmed this association.

b. Polymer The durable polymer which remains covering the stent after releasing the anti-stenotic drug has known to cause hypersensitivity and is suspected as the culprit for stent thrombosis and progressive or late re-stenosis (please see the above discussion under 'Histopathology'.

c. Matrix Metalloproteinases Circulating matrix metalloproteinase(MMP) have been associated with ISR as they play significant roles in migration of vascular smooth cells and matrix remodelling during healing post-stenting. Elevated levels of MMP-9 at baseline and MMP-2 and MMP-9 levels 24 hours post-percutaneous coronary intervention have proven to be strongly associated with the development of ISR following DES implantation. On the contrary, low and near normal levels of MMP-2 and MMP-9 were associated with a lack of a significant re-stenotic response.

d. Genetic factors As mentioned above can play a significant role in the inflammatory response, which can result in ISR.

B) Arterial Factors

1. Wall Shear Stress

The laminar flow of blood is a well known phenomenon, where the blood flows the fastest at the vessel center (or at the carina of a bifurcation) and slowest when it flows closest to the vessel wall (or at the ostium of a bifurcation). Hence, regions of low shear stress lead to accumulation of biological mediators, which promote atherosclerosis or neointimal formation. Adapting this principle would mean that a divider or the formation of a new carina should reduce the incidence of ISR. Kim at al demonstrated reduced occurrence of ISR by 'shotgun stenting' (i.e., simultaneous V-stenting with the formation of a new carina in the left main stem or other suitably sized vessels) but conversely, Stinus et al demonstrated increased target lesion revascularization rate with V-stenting when compared with the 'Crush' technique. This disparity can have a couple of explanations: 1) lesion location 2) post stent lumen size 3) reference lumen size 4) exposure of the new carina (formed by stent struts) for thrombus formation.

The presence of patchy areas of low shear stress within stented segments, secondary to local geometric factors (such as angulation or curvature) can predispose towards increased neointimal formation and hence ISR. This was demonstrated by Papafaklis et al on 6 month follow-up with BMS and Paclitaxel DES. However, this was not seen with Sirolimus DES. This difference can be explained by the differing pharmacological mode of action and shorter drug-release kinetics.

2. Progression of Atherosclerosis within a Stented Segment If the necrotic plaque within the stent progresses or if a lipid core plaque at the stent edge is not covered completely, then these lesions can progress to cause ISR or even thrombus formation.

3. "Thrombostenosis" Phenomenon Oikawa et al was the first to describe the intriguing theory in which chronic thrombus formation may play an integral role in the development of ISR.

4. Vessel Remodelling The implantation of DES in vessels that have already undergone positive remodelling secondary to large plaque burden ("Glagov" phenomenon) have an increased risk for ISR.

C) Stent Factors

1. Polymer Release Characteristics In one the earlier trials involving Paclitaxel, they found that the duration of the drug release had a far greater impact on the inhibition of neointimal proliferation than the actual drug dose delivered. Several molecular biological studies have indicated that the genetic mechanisms responsible for the inflammatory proliferative response remain potentially active for a period of 21 days after vessel wall injury. These polymer release kinetics could explain some of the disparities in the ISR among different DES. It seems clear that when the antistenotic drug is delivered at a set threshold dose for a sustained prolonged period of time may be required to control the inflammatory proliferative response and hence reduce the occurrence of ISR.

2. Non-uniform Drug Distribution

a. Ideally, the anti-stenotic drug should be delivered in a homogeneous fashion to the vessel wall (i.e., transmural and circumferential). However, due to local blood flow alterations (secondary to atherosclerotic plaque and calcified lesions), non-compliant vessels (i.e., calcified, tortuous), strut overlap, stent design, stent gap, vessel curvature, bifurcation lesions, ostial lesions, stent fracture and polymer damage can result in focal areas of the vessel receiving sub-optimal anti-stenotic drug.

b. Stent delivery in calcified lesions can result in stripping of the polymer base.

c. Finally, individual variations in metal to artery ratio and variability in drug elution can all contribute to stent re-stenosis.

3. Strut Thickness Thicker stent struts have been associated with increased risk of ISR with BMS and smaller vessels. This could be explained by the fact that the thinner stent strut would have a lesser dose of foreign body exposure to the vessel wall, hence reducing the intensity of the inflammatory proliferative response.

4. On- and Off-Label Use The STENT group is the largest, multicenter, prospective registry involving >15,000 patients, evaluating the late outcomes associated with DES implantation in the United States showed an almost 2 fold increase in target vessel revascularization (TVR) at 9 months for off-label use (ostial, left main stem, chronic total occlusions, saphenous venous graft, small or large vessels, multivessel, ST-elevation myocardial infarction, ISR lesions) when compared to on-label use (short de novo lesions measuring >2.5mm and <3.5mm for Sirolimus DES or <3.75mm for Paclitaxel DES). Also, when looking at the SYNTAX study, there were higher TVR rates in patients with highly complex 3-vessel or left main disease who underwent DES (which was an off-label use of DES). When analyzing the list of off-label uses, it seems that these patients have complex coronary lesions (with a higher SYNTAX score), more sicker and have more comorbidities which could easily influence TVR.

5. Polymer Disruption, Peeling and Crackling This can result in exposure of bare-metal areas, which has been demonstrated in bench studies involving both first and second generation DES using light or scanning electron microscopy. It is plausible that this bare metal exposure and stent regions in the absence of anti-stenotic drug can form a nidus for inflammatory and proliferative response. Seen more so in stent implantation in tortuous calcified lesions.

D) Mechanical Factors

1. Stent Underexpansion When the stent is deployed and it is not well expanded, it results in a vessel lumen size whose stent cross-sectional area is significantly smaller that of the reference lumen size or other cross-sectional areas within the stent, despite good apposition of the stent struts against the vessel wall. If the minimum stent area is smaller than the reference lumen size to begin with, then the same amount of neointimal formation would be of greater significance than in a scenario where the minimum stent area was larger. An excellent expansion is considered if the minimal cross-sectional area in the stent is >90% of the average reference cross-sectional area.

2. Stent Malapposition In malapposition, the stent struts are not apposed to the vessel wall (blood occupies the space between the stent struts and vessel wall). It is commonly seen with undersized stents, tortuous arteries and non-homogeneity in the reference arterial lumen diameter within the treated segment.

E) Stent Fracture

It is defined as complete or partial separation of a stent, which was contiguous on initial implantation. This commonly happens with non-uniform stent expansion in region with both very compliant and resilient vessel walls. IVUS is an accurate imaging modality to recognize this abnormality. Partial stent fracture is defined as the absence of atleast one third or 120 degrees of stent struts for at least 1 frame. Complete stent fracture is defined as the complete absence of stent struts within the stented segment for at least 1 frame. There have been various classification systems to represent the severity of stent fracture. The incidence of DES fracture has been reported from 1% to 8%. In fact, in the only randomized controlled trial reporting the incidence and outcomes of stent fracture in DES (LONG-DES II study), where they had a follow-up angiography, a 14% incidence of restenosis was observed.

There are 2 important mechanisms for stent fracture. Firstly, with excessive movements during myocardial contraction, significant tension is exerted at hinge points where the myocardial contractility forces act in opposing direction and result in fracture. Secondly, when a stent has a closed cell design (as seen with Sirolimus DES), the stent cannot dissipate the myocardial contractility pressures and hence the stent itself succumbs to this force and results in a fracture. Hence, with these principle mechanisms in mind; long stents, right coronary artery lesions, excessive tortuosity, angulation and torsion of the vessel, overlapping stents, saphenous vein graft lesions, tight calcified lesions that have been vigorously post dilated and expanded, myocardial bridges' sites and closed cell design stents (such as Sirolimus DES) predispose towards stent fracture.

F) Technical Factors

1. Barotrauma/Geographical Miss (GM) When a stent is deployed and if its fails to cover the sites of balloon injury or the atherosclerotic in its entirety, then these missed sites can become a nidus for re-stenosis. GM can be classified as longitudinal or axial. Longitudinal GM was defined as injured or diseased stenotic segment, which was not fully covered by the DES. Whereas, axial GM was defined as inadequate or overzealous sizing of the balloon, which results in effects similar to stent underexpansion/malapposition and uncontrolled vessel wall injury (increasing the stimulus for inflammation and proliferative response, respectively. Hence, the presence of injured or diseased segment not covered by the stent or balloon-artery ratio <0.9 or >1.3 was noted to have an increased association with target vessel revascularization and myocardial infarction at 1 year.

2. Stent Gap It is simply defined as a discontinuous coverage between 2 stents. A gap between 2 stents exposes the site of balloon injury as well the lack of anti-stenotic drug distribution (if it is in between 2 DES) will clearly increase the risk for re-stenosis

3. DES Deployment in a Clot-Laden Segment Logically, it seems that there is significant reduction in penetrance of anti-stenotic drug in regions, where the DES is deployed in clot-laden arterial segment. However, with the widespread use of glycoprotein-IIb/IIIa inhibitors, potent P2Y12 receptor antagonists and aspiration thrombectomy, these concerns do not translate into clinical events. This is demonstrated in a recent meta-analysis of 13 trials (n=7244) has shown significant benefits of DES over BMS in reducing TVR and recurrent MI (at the end of one year) in STEMI patients.

Clinical Presentation

In-stent restenosis (ISR) can be clinically silent, but majority of them present with recurrent symptoms of angina. The incidence of recurrent angina pectoris after a percutaneous coronary intervention (PCI) was reported in the past to be around 50% with a wide range. This number may have reduced as most the PCIs end up in a DES as opposed to a BMS. The positive predictive value of symptoms indicating a significant stenosis is as low as 60%. ISR is often thought to be a benign phenomenon since the process of neointimal formation and proliferation is of gradual onset and progressive in nature. Given the pathophysiology of coronary re-stenosis, it is thought that re-stenosis is a rare cause of acute myocardial infarction or death. However, there are several reports which have shown that ISR can present as an acute coronary syndrome. 26% to 53% and 3.5% to 20% of BMS ISR can present as unstable angina and myocardial infarction, respectively. Similarly, 16% to 66% and 1% to 20% of DES ISR can present as unstable angina and myocardial infarction, respectively. A highly stenotic ISR lesion can lead to an non-occlusive thrombus, which can result in an acute coronary syndrome. Also, patients with clinically silent re-stenosis can be identified on coronary cineangiograms when neighbouring plaques undergo rupture or intimal tear and present as an acute coronary syndrome. Sometimes, local plaque rupture or intimal tear can initiate an inflammatory process which can promote thrombosis of neighbouring stenotic lesions. Thus, it is important to thoroughly evaluate the patient as coronary re-stenosis can present as an acute coronary syndrome.

BMS ISR has been reported to occur usually after five and half months after stent implantation. The time frame for DES ISR presentation is not well-known with one study reporting a mean time duration of 12 months. Delayed restenosis is known to occur especially with DES. There have been reports, using intravascular ultrasound that have shown neointimal proliferation to occur even at 4 years after stent implantation. The exact reasons as to why this delayed neointimal proliferation occurs is not well known. Some of the suspected pathophysiological mechanisms are delayed healing response, persistent biological reaction caused by the drug present in the polymer, or a hypersensitivity reaction to the polymer and a possible a genetic predisposition. This eludes to the fact that the clinician should entertain the possibility of coronary re-stenosis in patients who present with recurrent angina about 2 years after the stent implantation.

Among the clinical predictors of coronary re-stenosis, diabetes mellitus continues to be a strong clinical predictor. In a study conducted by Singh et al., they found that patients with treated diabetes mellitus had a 45% higher risk of restenosis compared with nondiabetics. Interestingly, in their study they found that current smokers have less restenosis. This smoker's paradox has been described in the past. Some of the other predictors are increasing age, female sex and chronic renal disease and patients on hemodialysis. Angiographic and other predictors are listed under the other sections.

Related Chapters

References

  1. Dangas GD, Claessen BE, Caixeta A, Sanidas EA, Mintz GS, Mehran R (2010). "In-stent restenosis in the drug-eluting stent era". J Am Coll Cardiol. 56 (23): 1897–907. doi:10.1016/j.jacc.2010.07.028. PMID 21109112.
  2. Mehran R, Dangas G, Abizaid AS, Mintz GS, Lansky AJ, Satler LF; et al. (1999). "Angiographic patterns of in-stent restenosis: classification and implications for long-term outcome". Circulation. 100 (18): 1872–8. PMID 10545431.
  3. Farb A, Weber DK, Kolodgie FD, Burke AP, Virmani R (2002). "Morphological predictors of restenosis after coronary stenting in humans". Circulation. 105 (25): 2974–80. PMID 12081990.
  4. Moreno PR, Bernardi VH, López-Cuéllar J, Newell JB, McMellon C, Gold HK; et al. (1996). "Macrophage infiltration predicts restenosis after coronary intervention in patients with unstable angina". Circulation. 94 (12): 3098–102. PMID 8989115.
  5. Okamoto E, Couse T, De Leon H, Vinten-Johansen J, Goodman RB, Scott NA; et al. (2001). "Perivascular inflammation after balloon angioplasty of porcine coronary arteries". Circulation. 104 (18): 2228–35. PMID 11684636.
  6. Schiele TM (2005). "Current understanding of coronary in-stent restenosis. Pathophysiology, clinical presentation, diagnostic work-up, and management". Z Kardiol. 94 (11): 772–90. doi:10.1007/s00392-005-0299-x. PMID 16258781.
  7. Indolfi C, Curcio A, Chiariello M (2003). "Simvastatin reduces neointimal thickening after experimental angioplasty". Circulation. 107 (3): e25. PMID 12551884.

Template:WH Template:WS