|Home | About | Journals | Submit | Contact Us | Français|
Stroke is a leading cause of mortality and long-term morbidity. As a means for stroke prevention, an estimated 99,000 carotid endarterectomy procedures were performed in the USA in 2006. Traditionally, the degree of luminal stenosis has been used as a marker of the stage of atherosclerosis and as an indication for surgical intervention. However, prospective clinical trials have shown that the majority of patients with a history of recent transient ischemic attack or stroke have mild-to-moderate carotid stenosis. Using stenosis criteria, many of these symptomatic individuals would be considered to have early-stage carotid atherosclerosis. It is evident that improved criteria are needed for identifying the high-risk carotid plaque across a range of stenoses. Histological studies have led to the hypothesis that plaques with larger lipid-rich necrotic cores, thin fibrous cap rupture, intraplaque hemorrhage, plaque neovasculature and vessel wall inflammation are characteristics of the high-risk, ‘vulnerable plaque’. Despite the widespread consensus on the importance of these plaque features, testing the vulnerable plaque hypothesis in prospective clinical studies has been hindered by the lack of reliable imaging tools for in vivo plaque characterization. MRI has been shown to accurately identify key carotid plaque features, including the fibrous cap, lipid-rich necrotic core, intraplaque hemorrhage, neovasculature and vascular wall inflammation. Thus, MRI is a histologically validated technique that will permit prospective testing of the vulnerable plaque hypothesis. This article will provide a summary of the histological validation of carotid MRI, and highlight its application in prospective clinical studies aimed at early identification of the high-risk atherosclerotic carotid plaque.
Stroke is the leading cause of long-term disability as well as the third most common cause of mortality in the USA. Approximately 795,000 individuals experience a new or recurrent stroke each year in the USA, that is one person every 40 s . The estimated direct and indirect cost of stroke for 2009 is US$68.9 billion , and the projected costs for 2050 exceed $2.2 trillion, leading to a call for a greater focus on stroke prevention and improvement in the treatment of patients with acute stroke .
A number of randomized trials have documented the clinical benefit of carotid endarterectomy (CEA) for secondary stroke prevention among recently symptomatic patients with high-grade carotid stenosis [3-5]. In the North American Symptomatic Carotid Endarterectomy Trial (NASCET), patients with carotid atherosclerosis who had a recent history of transient ischemic attack (TIA) or nondisabling stroke were randomized to optimal medical therapy or CEA. In the subgroup with 70–99% stenosis, CEA was associated with an absolute risk reduction of 17% for ipsilateral stroke over 2 years .
The role of CEA in symptomatic patients with less than 70% carotid disease is less clear. In this more moderate stenosis group, CEA afforded an absolute risk reduction of only 6.5% for stroke over the following 5 years, compared with best medical therapy. As such, 15 CEA procedures would need to be performed to prevent one stroke over 5 years.
While results from NASCET and other prospective trials indicate a higher risk for stroke with severe carotid stenosis , it is noteworthy that the large majority of the subjects in these trials who presented with recent carotid territory ischemic events were found to have mild-to-moderate stenosis. The European Carotid Surgery Trial reported that 43.8% of the 3018 individuals with symptomatic carotid disease had less than 30% stenosis on angiography . NASCET reported that 61% of the 2226 recently symptomatic subjects had less than 50% carotid stenosis . Also of note, the risk for stroke in NASCET was similar amongst those with 50–69% stenosis (22.2% at 5 years) and those with 0–49% stenosis (18.7% at 5 years) . These findings suggest that severity of carotid stenosis is a poor discriminator of stroke risk amongst those with mild-to-moderate luminal narrowing.
Based on these observations, additional criteria have been sought to better identify patients most at risk for complications from carotid atherosclerosis. Wasserman, Virmani and colleagues have advocated looking beyond luminal narrowing to identify the high-risk carotid plaque. They and others have shown that measurement of stenosis, using the method described in NASCET, underestimates plaque burden [9-11]. This is likely related to the geometry of the carotid bulb, which is normally larger than the more distal internal carotid artery (Figure 1), as well as the phenomenon of expansive remodeling, originally described by Glagov (Figure 2) . Based on the geometry of the bulb, it is possible to have an eccentric plaque opposite the flow divider, despite measurement of a 0% stenosis, using the method described in NASCET (Figure 1).
Beyond simple plaque burden, a number of investigators have hypothesized that specific compositional features of the plaque also distinguish the high-risk, ‘vulnerable plaque’ from the stable, clinically silent lesion. Analysis of histological findings in CEA specimens has shown that fibrous cap (FC) rupture, intraplaque hemorrhage (IPH), large lipid-rich necrotic cores, erosions with overlying mural thrombus, plaque neovasculature and inflammatory cell infiltration are more commonly observed in plaques removed from previously symptomatic patients [13-19].
In summary, findings from the symptomatic carotid clinical trials indicate that the majority of patients with recent carotid territory ischemic events have mild (<30%) to moderate (30–69%) carotid stenosis. Additional plaque parameters, other than quantification of stenosis, are needed to identify the high-risk plaque amongst these individuals who are presumed to have early-stage carotid atherosclerosis.
Progress toward prospectively testing the ‘vulnerable plaque hypothesis’ has been hampered by the inability to accurately and reproducibly identify key plaque features that are believed to represent the high-risk lesion in vivo. Transcutaneous B-mode ultrasonography provides a relatively inexpensive method for imaging carotid plaque with high resolution. Studies dating back to the early 1980s have shown that plaques with an echolucent, heterogeneous or ulcerated appearance are associated with TIA or stroke [20-27]. Challenges for ultrasound include acoustical shadowing from calcification, anisotropic effects (where the appearance of the lesion can vary depending on the angle of insonation) [28-30], operator variability, lack of specificity for distinguishing lipid core from IPH and modest reader reproducibility [27,31-34]. CT and PET are promising modalities to quantify atherosclerotic carotid artery lesion size and compositional features, particularly calcification (CT) [35-39] and vessel wall inflammation (PET) [40-43]. Advantages of MRI include its superior specificity for characterizing tissue composition, image generation without ionizing radiation and extensive histological validation for characterizing carotid atherosclerosis. Multiple centers have shown that MRI can reliably identify FC status [44-48], plaque composition [49-58], neovasculature and vascular wall inflammation [59,60] using histology as the gold standard. This has been achieved via the development of custom designed surface coils that result in significant improvement in the signal-to-noise ratio, and specialized multicontrast-weighted imaging sequences that include bright-blood time-of-flight (TOF) as well as pre- and postgadolinium contrast-enhanced black-blood imaging, which provide submillimeter (~0.6 mm) in-plane resolution. Details of hardware development, pulse-sequence design and MRI criteria for carotid plaque characterization have been previously published [61-67].
Using a multicontrast-weighted protocol including a 3D TOF bright-blood imaging technique, Yuan et al. described the MRI appearance of intact/thick FC as a continuous hypointense band near the bright lumen on 3D-TOF images, and a smooth luminal surface. Plaques where the hypointense band could not be visualized were categorized as having an intact/thin FC. FC rupture was identified by the absence or discontinuity of the hypointense band, juxtaluminal hyperintense signal in the TOF and T1-weighted (T1W) images (consistent with recent hemorrhage), and/or an irregular lumen surface. In a study comparing in vivo MRI with histology in patients scanned prior to CEA, Yuan found a high level of agreement between the magnetic resonance (MR) findings and the histological state of the FC, with a k value of 0.83 (95% CI: 0.67–1.0) and a weighted k value of 0.87 . The sensitivity and specificity for identifying a thin or ruptured cap was 81 and 90%, respectively . Figure 3 illustrates the appearance of the FC on 1.5 T MRI, and Figure 4 demonstrates FC rupture with ulceration and thrombus formation on 3 T MRI, with corresponding histology. Figure 5 demonstrates cap rupture with ulceration in the common carotid artery on 3 T MRI.
A more recently published study by Cai et al.  demonstrated the utility of gadolinium-based contrast-enhanced MRI for increasing the conspicuity of the FC and the lipid-rich necrotic core (LRNC), permitting more reliable quantification of FC and LRNC dimensions. A total of 108 cross-sectional locations with intact FCs from 21 arteries were matched between MRI and the excised histology specimens. Quantitative measurements of FC length along the lumen circumference, FC area and LRNC area were collected from contrast-enhanced MR images and histology sections (Figure 6). Blinded comparison of corresponding MRI and histology slices showed moderate to good correlation for length (r = 0.73, p < 0.001) and area (r = 0.80, p < 0.001) of the intact FC. The mean percent LRNC areas ([LRNC area divided by the wall area] × 100%) measured by contrast-enhanced MRI and histology were 30.1 and 32.7%, respectively, and were strongly correlated across locations (r = 0.87, p < 0.001). Intraobserver reproducibility was excellent for LRNC area, FC area and FC length (intraclass correlation coefficient = 0.87, 0.72 and 0.80, respectively). Interobserver reproducibility was also excellent for LRNC area, FC area and FC length (intraclass correlation coefficient = 0.89, 0.78 and 0.81, respectively).
Intraplaque hemorrhage, deep within the core of the lesion, and hemorrhage/thrombus near or on the luminal surface may differ in etiology and clinical implications. In a study of 26 patients scheduled for CEA, Kampschulte et al. compared preoperative carotid MRI findings to the patients’ matched histology to determine whether MRI can distinguish between deep IPH and juxtaluminal hemorrhage/mural thrombus . Hemorrhages were identified using previously established MRI criteria (type I = more recent hemorrhage: hyperintense on TOF and T1W images; type II = older hemorrhage: hyperintense on TOF, T1W, proton density-weighted and T2W images)  and their locations were differentiated between intraplaque and juxtaluminal. Corresponding histology was used to confirm the MR findings. Matched sections (n = 190) contained 144 areas of hemorrhage by histology, of which MRI correctly detected 132 areas. The sensitivity and specificity for MRI to correctly identify cross-sections containing hemorrhage was 96 and 82%, respectively. Furthermore, MRI was able to distinguish juxtaluminal hemorrhage/thrombus from IPH with an accuracy of 96%.
Moody and colleagues have shown that the contrast between plaque hemorrhage and other plaque components can be improved using an inversion-prepared rapid 3D gradient-echo sequence, also known as MP-RAGE (Figure 7) . In a study of 63 patients who underwent an MRI scan prior to CEA, the authors reported a sensitivity and specificity of 84%, and very good reproducibility (inter- and intraobserver κ = 0.75 and 0.9, respectively).
In the 1930s, Winternitz reported that neovasculature may be involved in the pathogenesis of atherosclerosis . More recently, O’Brien noted that neovasculature within plaques may represent a pathway for the recruitment of macrophage infiltration, and that the endothelial cells lining these microvessels are a site of inflammatory activation . Work by Galis, Libby and Nikkari have shown that macrophages, typically found in the shoulder regions adjacent to the FC, express matrix metalloproteinases that can result in the weakening of the FC, predisposing it to rupture [71-73]. More recently, Levy et al. have suggested that plaque neovasculature may play an important role in the pathogenesis of IPH .
Work by a number of investigators have demonstrated differential enhancement of carotid plaque tissues using gadolinium-based MRI contrast agents [50,75,76]. They found that strong enhancement generally suggests the presence of a highly permeable vascular supply within the plaque (neovasculature) and loose extracellular matrix for contrast agent uptake. As neovasculature and increased endothelial permeability are both associated with plaque inflammation [70,77-80], gadolinium enhancement of the vessel wall has been hypothesized to be a marker of the vascular wall inflammation. To probe this hypothesis further using quantitative analyses, Kerwin et al. used dynamic contrast-enhanced MRI to measure the rate of uptake of gadolinium-based contrast, characterized by the transfer constant Ktrans, and compared these measurements to histological measurements of plaque composition and inflammation. The parameter Ktrans is well known in oncology, where it has been used to characterize tumor blood supply and permeability .
To measure Ktrans, repeated MRI measurements were made over short intervals to observe the dynamics of enhancement in the tissue of interest and in a reference arterial lumen (the ‘arterial input function’). The arterial input function and tissue-enhancement curves were then fit by a parametric model of contrast agent kinetics. Kerwin utilized the equation:
where Ct, and Cp are the contrast agent concentrations in the tissue (total), and blood plasma, respectively, and υp is the partial volume of blood plasma. This is the standard kinetic model  with a vascular term and ignoring reflux. Concentrations were assumed to be proportional to intensity change. The results of this modeling are parametric maps of υp and Ktrans in the plaque that not only quantify the amount of enhancement, but also the rate of enhancement.
Using this dynamic MRI acquisition protocol, Kerwin measured υp and Ktrans in 30 patients scheduled to undergo CEA . The excised specimens were then histologically analyzed to measure neovasculature and macrophage content, expressed as percent area (% neovasculature = [neovasculature area divided by plaque area] × 100%). Measurements of υp correlated with neovasculature content (Figure 8A; r = 0.68, p < 0.001). Ktrans also correlated with neovasculature (Figure 8B; r = 0.71, p < 0.001) and macrophage content (Figure 8C; r = 0.75, p < 0.001). Interestingly, there was a negative correlation between Ktrans and high-density lipoprotein (HDL)-cholesterol levels (r = -0.66, p < 0.001). Ktrans was also noted to be significantly higher amongst cigarette smokers compared with nonsmokers (mean: 0.134 vs 0.074 min-1; p = 0.01). Both low HDL and smoking have been found to be proinflammatory stimuli for atherosclerosis [83,84].
In summary, carotid MRI is a histologically validated tool that can identify the key features that are believed to characterize the vulnerable plaque, including FC status, plaque composition, neovasculature and vascular wall inflammation. Furthermore, MRI is ideally suited for serial assessment of temporal changes in the lesion in a noninvasive, nondestructive fashion. Therefore, MRI provides a critical tool to test the vulnerable plaque hypothesis in vivo.
In a prospective, observational study of 154 subjects with 50–79% carotid stenosis who were asymptomatic at the time of enrollment, Takaya et al. tested the hypothesis that specific carotid plaque features are associated with a higher risk of subsequent TIA or stroke . Following their baseline carotid MRI examination, subjects were called every 3 months to identify symptoms of new-onset TIA or stroke; 12 cerebrovascular events (four strokes and eight TIAs) that were judged to be carotid related occurred during a mean follow-up period of 38.2 months. Cox regression analysis demonstrated significant associations between ischemic events and presence of a thin or ruptured FC (hazard ratio: 17.0; p < 0.001), IPH (hazard ratio: 5.2; p = 0.005) and larger mean necrotic core area (hazard ratio for 10 mm2 increase: 1.6; p = 0.01) in the carotid plaque (Figure 9A & B).
These findings were corroborated in a more recently published study by Singh et al . A total of 91 initially asymptomatic men with 50–70% stenosis were followed for a mean period of 25 months. They reported that of the six cerebrovascular events that occurred, 100% corresponded to arteries with IPH present at baseline (hazard ratio: 3.6; p < 0.001) .
The value of MRI in patients who had a recent history of TIA or stroke and moderate carotid stenosis was recently demonstrated by Altaf et al. . Amongst 64 subjects with 30–69% carotid stenosis and a recent carotid territory ischemic event, Altaf performed baseline carotid MRIs to identify IPH and followed the subjects for the development of subsequent TIA or stroke. Of all the index arteries demonstrated, 39 (61%) demonstrated IPH on baseline MRI. Median follow-up was 38 months. A total of 14 ipsilateral ischemic events (nine TIAs and five strokes) were observed during follow-up. Out of the 14 events, 13 occurred ipsilateral to carotid arteries with IPH (hazard ratio: 9.8; 95% CI: 1.3–75.1; p = 0.03).
The small number of ischemic events in these studies [85-87] precluded multivariate analyses. Nonetheless, these findings offer compelling prospective evidence of the potential of carotid MRI for defining the high-risk, vulnerable plaque, and provide a foundation for the design of larger, multicenter studies.
As a noninvasive imaging modality, MRI is ideally suited for serial examination of plaque features that may play a critical role in the pathogenesis of high-risk lesions of atherosclerosis. In a histopathology study of excised coronary arteries, Kolodgie, Virmani and colleagues suggested that IPH may represent a potent atherogenic stimulus by contributing to the deposition of free cholesterol, macrophage infiltration and enlargement of the necrotic core . In a case–control study of 29 subjects participating in a longitudinal, serial MRI progression study, Takaya et al. tested the hypothesis that IPH, as detected by high-resolution MRI, was associated with greater progression in both necrotic core and plaque volume . The volume of wall, lumen, necrotic core and IPH were measured at baseline and follow-up. Carotid arteries with IPH on baseline examination demonstrated markedly accelerated rates of progression in wall volume (6.8% with IPH vs -0.15% without IPH; p = 0.009) and LRNC volume (28.4% with IPH vs -5.2% without IPH; p = 0.001) over the course of 18 months (Figure 10). Furthermore, those with IPH at baseline were much more likely to have new plaque hemorrhages at 18 months compared with controls (43 vs 0%; p = 0.006). Findings from this study strongly suggest that hemorrhage into the carotid atherosclerotic plaque accelerates plaque progression within a relatively short period of 18 months.
Underhill et al. found a similar pattern in individuals with subclinical, earlier-stage carotid atherosclerosis . In a prospective, longitudinal MRI study of 67 asymptomatic subjects with 16–49% stenosis, IPH was associated with accelerated progression in carotid wall volume compared with lesions without IPH (44.1 ± 36.1 vs 0.8 ± 34.5 mm3 per year; p < 0.001). Underhill also found that IPH altered the pattern of arterial remodeling. Lesions without IPH demonstrated outward, expansive remodeling with preservation of luminal dimension, as originally described by Glagov et al. . However, in carotid arteries with IPH, these compensatory mechanisms appeared to be overridden by rapid expansion of the lesion, and was associated with luminal narrowing (-24.9 ± 21.1 mm3 per year; p = 0.002).
As noted earlier, MRI identification of FC rupture is highly associated with carotid territory ischemic events [85,91]. In a prospective, serial MRI study of 85 subjects with 50–79% stenosis and no luminal surface disruption at baseline, Underhill et al. examined the clinical and baseline carotid MRI plaque features that were associated with new FC disruption on follow-up MRI . They found that the size of the LRNC at baseline was the strongest classifier for development of a new surface disruption at the 36 month follow-up scan (area under the curve [AUC] = 0.95). Presence of IPH was also a statistically significant, but weaker, classifier of new surface disruption (AUC = 0.73).
Oikawa et al. examined the relationship between the extent of adventitial vasa vasorum, as measured by Ktrans on baseline dynamic contrast-enhanced MRI, and the development of new FC rupture at the 36 month follow-up . Amongst 30 arteries evaluated, baseline mean Ktrans was significantly higher in arteries with FC rupture and/or IPH at baseline (n = 14) compared with those with intact surface without IPH (n = 16) (0.13 ± 0.005/min vs 0.10 ± 0.004/min, respectively; p < 0.001). Furthermore, amongst 23 arteries with an intact FC on the initial MRI, baseline mean Ktrans was significantly higher in arteries that developed MRI evidence of new FC rupture at 3 years (n = 4) compared with those with intact surface at follow-up (n = 19) (0.14 ± 0.009/min vs 0.10 ± 0.004/min, respectively; p = 0.004). A total of 50% of the arteries with a baseline mean Ktrans > 0.114/min developed FC rupture during follow-up (AUC = 0.91). These results suggest a potential role of adventitial neovasculature in the pathogenesis of plaque disruption. However, these preliminary findings require further investigation in larger studies.
MRI studies have shown that presumptive high-risk plaque features, such as IPH and FC rupture, are commonly identified in carotid arteries with minimal-to-moderate stenosis [94,95]. In a review of 260 carotid MRI examinations performed in asymptomatic subjects, the prevalence of plaques with MRI evidence of IPH or FC rupture across a range of luminal stenoses was assessed by Saam et al. . Up to a third of subjects with asymptomatic 50–79% stenosis and approximately 10% with 16–49% stenosis have evidence of cap rupture or IPH (Figure 11).
In a more recent study of subjects undergoing gadolinium contrast-enhanced magnetic resonance angiography (CE-MRA) and high-resolution carotid MRI at 3 T, Dong et al. found a surprisingly high prevalence of complex plaque features in arteries with 0% stenosis . A total of 72 individuals with more than 50% carotid stenosis in at least one carotid artery by duplex ultrasonography were recruited for MRI of their bilateral carotid arteries. For each artery, the percent wall volume (wall volume/[lumen volume + wall volume] × 100%) and the prevalence of LRNC, calcification, IPH and FC rupture were recorded. Of the 144 arteries available for analysis, 133 had interpretable image quality, and 36.1% of the remaining arteries had a 0% stenosis on CE-MRA, as measured using NASCET criteria. Amongst arteries found to have a 0% stenosis by CE-MRA, the mean percent wall volume was 43.0 ± 6.9% with a range from 31.6 to 60.1%. LRNC was present in 67.4% (31 out of 46) of arteries, calcification was present in 65.2% (30 out of 46), IPH was present in 8.7% (4 out of 46), and FC rupture was present in 4.3% (2 out of 46). While selection criteria for the study limit extrapolation of these results to the general population, the findings confirm that angiography underestimates carotid plaque burden (Figure 12) [9,10] and is ineffective in detecting the presence of complex plaque.
The clinical studies reviewed herein demonstrate the potential prognostic value of carotid MRI for subsequent TIA or stroke in asymptomatic and recently symptomatic individuals with moderate carotid stenosis. Furthermore, MRI may play a valuable role in examining mechanisms involved in the pathogenesis of the high-risk carotid plaque. While the results from these initial studies are promising, the small number of events precluded multivariate analysis of the data, and highlights the need for larger, multicenter studies.
Additionally, initial findings suggest the need for imaging criteria in the form of a comprehensive plaque score that would provide greater positive predictive value for future events. In the study by Altaf et al. , IPH was equally prevalent in the contralateral (nonindex) carotid artery, yet there was only one TIA and no strokes referable to the nonindex side during follow-up. This suggests that factors other than the sole presence of IPH may also be important, such as FC status, size of the LRNC and IPH, the location of IPH (juxtaluminal or deep within the plaque) and degree of neovasculature. Larger studies are needed to test the hypothesis that specific baseline plaque features, independently or in combination, are associated with an increased risk for future ischemic events.
Validation of a comprehensive plaque score in prospective observational studies will provide the foundation for future randomized clinical trials. These trials will be needed to assess whether new plaque imaging-based selection criteria, by identifying individuals at higher risk than those identified by current stenosis-based criteria, will yield a greater absolute risk reduction that will be sufficient to justify carotid surgery or stenting in patients with moderate carotid stenosis.
Finally, the development of novel, molecular imaging techniques will advance the MRI field to the next critical level. The current clinically available MRI techniques described in this article provide important information regarding the structure and composition of human carotid atherosclerosis in vivo. Furthermore, dynamic contrast-enhanced MRI shows promise for providing an indirect measure of plaque neovasculature and the inflammatory activity of the plaque. The development of probes that package MRI contrast agents with targeted lipid-based nanoparticles [97-99] or HDLs  show great promise for providing a direct measure of plaque activity and function at a molecular level .
Carotid MRI has been extensively validated and provides quantitative information regarding the morphology and composition of human carotid atherosclerosis in vivo. Therefore, it provides an essential tool that will allow prospective studies to test the vulnerable plaque hypothesis.
Currently, there is a lack of consensus in the management of recently symptomatic patients with moderate carotid stenosis. Furthermore, improved methods of plaque imaging have documented that stenosis severity underestimates carotid plaque burden and plaque complexity, including those with angiographically normal appearing arteries. It is critical that we develop better plaque imaging-based methods for stroke risk stratification so that individuals with stable plaques will be spared from unnecessary surgery or stenting, and individuals with unstable, high-risk lesions who would be appropriately referred.
Finally, a better understanding of the characteristics of the vulnerable plaque will provide a foundation for further research on the pathogenesis of high-risk lesions, and perhaps lead to development of novel pharmacological therapy.
The authors wish to acknowledge and are extremely grateful to their collaborators and colleagues, whose guidance, support and effort has been invaluable: N Balu, B Chu, M Ferguson, WS Kerwin, HR Underhill, D Xu, V Yarnykh, SC Cramer, G Garden, GP Jarvik, K O’Brien, NL Polissar, R Ross, S Schwartz, R Virmani, J Cai, G Canton, L Dong, A Kampschulte, F Li, M Oikawa, H Ota, T Saam, N Takaya, W Yu, X Zhao, H Chen, F Liu, J Wang, W Hamar, D Hippe, C Isaac, D Jensen, Z Miller, R Small, R Smith and T Zhu.
Financial & competing interests disclosure
The authors have received grant funding in the past from the NIH (R01 HL073401, P01 HL072262, R01 HL61851 and R01 HL56874), which has supported some of the work reviewed in this manuscript. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.
No writing assistance was utilized in the production of this manuscript.
Papers of special note have been highlighted as: