Background Non-alcoholic fatty liver disease (NAFLD) is a well-established risk factor for cardiovascular disease, with ethnic and regional differences noted. With the recent surge of research within this field, we re-examine the evidence associating NAFLD with subclinical atherosclerosis, and investigate potential regional differences.
Methods This is a systematic review and meta-analysis. PubMed and EMBASE were systematically searched for publications from January 1967 to July 2020 using standardised criteria. Original, observational studies investigating the association between NAFLD and either carotid intima-media thickness (CIMT) and/or coronary artery calcification (CAC) were included. Key outcomes included differences in mean CIMT, the presence of increased CIMT, the presence of CAC and the development/progression of CAC. Pooled ORs and pooled standard differences in means were calculated using random-effects models. Between-study heterogeneity was quantified using the Q statistic and I². Subgroup analyses stratified by region of study (Asian vs Western) were also conducted.
Results 64 studies involving a total of 172 385 participants (67 404 with NAFLD) were included. 44 studies assessed the effect of NAFLD on CIMT, with the presence of NAFLD associated with increased CIMT (OR 2.00, 95% CI 1.56 to 2.56). 22 studies assessed the effects of NAFLD on CAC score, with the presence of NAFLD associated with the presence of any coronary calcification (OR 1.21, 95% CI 1.12 to 1.32), and the development/progression of CAC (OR 1.26, 95% CI 1.04 to 1.52). When stratified by region, these associations remained consistent across both Asian and Western populations (p>0.05). The majority (n=39) of studies were classified as ‘high quality’, with the remaining 25 of ‘moderate quality’.
Conclusions There is a significant positive association between various measures of subclinical atherosclerosis and NAFLD, seen across both Western and Asian populations. These results re-emphasise the importance of early risk evaluation and prophylactic intervention measures to preclude progression to clinical cardiovascular disease in patients with NAFLD.
- metabolic syndrome
Data availability statement
Data are available upon reasonable request.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
What is already known about this subject?
Non-alcoholic fatty liver disease (NAFLD) is a significant, independent risk factor for cardiovascular disease (CVD), with recent evidence positing this association to extend to the preclinical stages of CVD. Previous meta-analyses have quantified positive associations between NAFLD and subclinical atherosclerotic markers before, though the majority of included studies were published before 2016. The last 5 years, however, has experienced a large surge of research in this field, especially within large Asian populations that have not been included in previous meta-analyses. Ethnic and regional differences in the associations between NAFLD and subclinical atherosclerosis have been suggested within individual studies, but have yet to be synthesised across the available literature.
What does this study add?
This meta-analysis serves as a timely update of the existing literature, incorporating the results of over 21 new studies comprising over 100 000 participants (~50 000 with NAFLD) from both Western and Asian regions. The results reinforce the significant positive association between NAFLD and subclinical atherosclerosis (as defined by increased carotid intima-media thickness and coronary artery calcification scores), and further confirm these associations to be consistent across both Western and Asian populations. Lastly, this is the first meta-analysis to demonstrate that the associations between NAFLD and subclinical atherosclerosis are not just crosssectional but also longitudinal.
How might this impact on clinical practice?
This study highlights that NAFLD serves as an important atherogenic risk factor in both Western and Asian populations, and reemphasises the role of early risk evaluation and prophylactic intervention measures to preclude progression to clinical CVD in NAFLD. By confirming a longitudinal association between NAFLD and subclinical atherosclerotic markers, these results also provide potential insight into the causal relationship between NAFLD and subclinical atherosclerosis.
Non-alcoholic fatty liver disease (NAFLD) encompasses a wide spectrum of pathological hepatic conditions ranging from simple steatosis to non-alcoholic steatohepatitis, and may ultimately progress to advanced fibrosis, cirrhosis, and end-stage liver disease.1–3 Over the last 20 years, NAFLD has become the leading cause of chronic liver disease, with an estimated 1 billion people affected worldwide.4 NAFLD has increasingly been recognised as the hepatic manifestation of the metabolic syndrome (MetS), and is one facet of a multisystem disease, with close relations to abdominal obesity, type 2 diabetes mellitus (T2DM), insulin resistance and hyperlipidaemia.5–7
Cardiovascular disease (CVD) is the leading cause of mortality in patients with NAFLD, with a large body of evidence demonstrating NAFLD to be a significant, independent risk factor for CVD.2 3 5 8 It is now widely hypothesised that NAFLD is not merely a marker of CVD, but may be actively involved in CVD pathogenesis.3 9–11 This association extends to preclinical CVD, with recent work identifying NAFLD as a risk factor for early subclinical atherosclerosis, and as a strong independent predictor of incident CVD.3 8 12 This holds highly important implications for the screening and early evaluation of CVD in patients with NAFLD. Carotid intima-media thickness (CIMT) and coronary artery calcification (CAC) are the two most established and widely studied surrogate measures of subclinical atherosclerosis, and a growing body of literature has investigated this relationship between NAFLD and CAC/CIMT over the last decade.3 12 13
NAFLD is no longer considered a disease only prevalent in affluent Western countries, with rapidly growing rates of NAFLD reported within Asia in particular.4 14 Ethnic and regional differences in NAFLD prevalence, severity and outcomes have been identified between Western, Hispanic and Asian populations,15 16 and have been attributed to factors including lifestyle, environment, insulin resistance, body composition (adipose distribution and muscle bulk) and genetics.14 15 17 18 Asian populations are especially susceptible with cardiometabolic complications such as NAFLD seen to develop within a much shorter period, within younger patient populations and in those with lower body mass index.18 19 These disparities may possibly extend to differential associations between NAFLD and subclinical atherosclerosis.20–22
Previous meta-analyses have quantified the associations between NAFLD and subclinical atherosclerotic markers before, with the majority of included studies published before 2016.13 23–25 However, the last 5 years has experienced a large surge in research within this area, especially within large Asian populations that have not been reported in previous meta-analyses.26–35 We aim to evaluate the relationship between NAFLD and subclinical atherosclerosis including these updated studies, and to further investigate potential regional differences in these associations.
This meta-analysis was conducted and reported according to the Meta-analysis Of Observational Studies in Epidemiology statement36 and was registered in the International Prospective Register of Systematic Reviews (registration number: CRD42020204784).
A comprehensive literature search was performed via the MEDLINE and EMBASE databases to identify potentially relevant publications in the English language, with a date range from January 1967 to July 2020. The databases were systematically searched using a combination of the following keywords linked with appropriate Boolean logic: (Fatty Liver OR NAFLD OR Hepatic Steatosis OR Non-alcoholic fatty liver disease) AND ((subclinical atherosclerosis OR Preclinical atherosclerosis) OR (Coronary calcium OR Calcium Score OR Coronary Calcification) OR (“Carotid Intima-media thickness” OR CIMT OR IMT OR “‘intima media thickness”’)). Relevant references identified from the bibliographies of pertinent articles or review papers were also retrieved.
Eligibility (inclusion and exclusion) criteria
The eligibility criteria was based on the PICOS framework as recommended by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement.37
Participants: studies had to be conducted on adult participants. Studies conducted on ‘special populations’ including adolescent/paediatric populations, and those defined by additional pathologies such as HIV, severe CVD or liver transplants were rejected. Populations with existing metabolic conditions such as MetS and diabetes mellitus were accepted.
Exposures (intervention): studies had to have a defined exposure of ‘NAFLD’ or ‘fatty liver’ or ‘hepatic steatosis’, as diagnosed by either ultrasound (US), liver biopsy, CT, magnetic resonance spectroscopy (MRS) or Fatty Liver Index.
Outcomes: study outcomes had to report on either the (1) presence (cross-sectional) of CAC (CAC score >0), (2) progression (longitudinal) of CAC score, and/or (3) on CIMT. The presence of calcified coronary artery plaques was accepted as a measure of CAC score>0. Studies had to specify how CAC and CIMT were recorded and defined, and also had to quantitatively assess the association between NAFLD and CAC/CIMT, respectively, either via logistic regression for categorical outcomes or via comparison of means techniques (t-test/analysis of variance (ANOVA)) for continuous outcomes.
Comparison: studies had to include a ‘healthy’ control group of participants without NAFLD, preferably from the same population as the exposure group.
Study design: we included observational studies (cross-sectional, case–control, retrospective, prospective), which reported quantitative outcomes. Descriptive studies, reviews and studies on animals were excluded. Studies with sample sizes <50 were also excluded.
Using our search strategy, a total of 1007 titles were initially identified. Two authors (MYZW and JJLY) assessed the titles independently according to the predefined inclusion and exclusion criteria. Studies were first screened by title and abstract. The full-text articles deemed potentially relevant were then obtained and systematically included after detailed examination. The following data were extracted: (a) study: year, region, design; (b) patients: mean age, gender, sample size; (c) method of NAFLD evaluation: US, CT, MRS, liver biopsy or composite index; (d) outcomes: outcome type (CIMT or CAC) and method of outcome definition; (e) analysis: statistical techniques used, primary outcomes (mean±SD, ORs with 95% CIs), confounders adjustment.
For studies reporting multiple multivariable-adjusted models, we extracted those reflecting the greatest degree of control for potential confounders. Any discrepancies in data quantification were resolved by discussion among the investigators.
Study quality evaluation
The quality of observational studies was assessed using a modified version of the Newcastle–Ottawa Scale (NOS) for cohort and cross-sectional studies.38–40 The NOS awards a maximum of 9 stars to assess quality based on three main aspects: (a) the selection and representativeness of the participants (maximum 4 stars), (b) the comparability of groups (maximum 2 stars), and (c) the ascertainment of exposure (for case–control) or outcome (for prospective and cross-sectional) (maximum 3 stars). Following previous reviews, studies assigned 0–4, 5–7, and ≥8 stars were considered as low, medium and high quality, respectively.41–43
Data synthesis and statistical analysis
Outcomes were broadly grouped according to four main categories:
Differences in mean CIMT (continuous).
Presence of increased CIMT (categorical).
Presence of CAC (categorical).
Development/progression of CAC (categorical, longitudinal).
All outcomes were pooled using DerSimonian-Laird random-effects model. The continuous and categorical outcome was reported as pooled standard differences (Std Diff) in means and ORs with 95% CI. We further conducted subgroup analysis to look into regional differences between Asian versus Western populations. We defined ‘Western’ studies to comprise of studies conducted in North America, Europe and Australia, while ‘Asian’ studies comprised of those conducted in South Asian, East Asian and Southeast Asian countries. Lastly, additional subgroup analysis on the Std Diff in mean CIMT within the subset of participants with diabetes was conducted.
The heterogeneity of pooled estimates between studies was quantified using the Q statistic and I². A value of I² of 0%–25% indicates no heterogeneity, 26%–50% low heterogeneity, 51%–75% moderate heterogeneity and 76%–100% high heterogeneity. Funnel plots and Egger’s regression test were used to assess publication bias. P<0.05 was considered as statistical significance.
All statistical analyses were conducted using the Comprehensive Meta-Analysis Software V.3.3.
Search strategy and description of studies
The initial search yielded 1007 potentially relevant titles, where 835 articles were excluded on the basis of title and abstract screen. A total of 172 titles underwent full-length review, of which 108 were further excluded (figure 1). A final total of 64 studies, involving 67 404 patients with NAFLD and 104 981 controls were included in the meta-analysis. Tables 1–3 describe the detailed characteristics of the included studies, grouped by study outcome. These included studies were carried out in Asia (n=32), Western Europe (n=15), the Middle East (n=10) and America (n=7; North America: 6, South America: 1). Sixty studies were cross-sectional and four were prospective cohort studies.
Measurement of exposures and outcomes
The presence of NAFLD was largely determined by US (n=46), with other studies using CT (n=8), biopsy (N=8), Fatty Liver Index (n=1) and MRS (n=1). Twenty-two studies investigated the effects of NAFLD on CAC score, with one study using the presence of calcified coronary artery plaques as a proxy for CAC >0. Forty-four studies investigated the effects of NAFLD on CIMT score. CIMT was assessed via B-mode US of bilateral carotid arteries, with majority of studies (n=18) commonly averaging the mean CIMT over six measurements (three on each carotid artery).
Tables 1–3 and online supplemental table 1 detail the NOS risk of bias evaluation for the various studies. Of the 60 cross-sectional studies, the majority (n=35) were classified as ‘high quality’ (≥8 stars) with the remaining 25 classified as ‘moderate quality’ (5–7 stars). All four prospective studies were classified as ‘high quality’.
Effect of NAFLD on CIMT
Figures 2 and 3 summarise the studies which investigated the effects of NAFLD on CIMT. Forty-four studies, with a total of 41 189 individuals, assessed the effect of NAFLD on CIMT. Thirty-nine studies investigated the mean differences in CIMT between NAFLD and controls,22 30 33 35 44–79 while 13 studies used logistic regression to quantify the associations between NAFLD and an ‘increased CIMT’.22 28 44 45 59 61 62 74 79–83 Increased CIMT was defined as >0.8 mm in six studies, >1.0 mm in two studies and via other stratification methods in the remaining five studies.
Compared with participants without NAFLD, the presence of NAFLD was significantly associated with an increased CIMT, with a pooled OR of 2.00 (95% CI 1.56 to 2.56, Pheterogeneity<0.001, I2=81.8%, figure 2). Likewise, subjects with NAFLD had a higher mean CIMT than subjects without, both across studies which adjusted for confounders (pooled Std Diff in means: 1.17, 95% CI: 0.49 to 1.85, figure 3B), and in studies which compared unadjusted means (pooled Std Diff in means: 0.68, 95% CI: 0.44 to 0.91, figure 3A). For all CIMT outcomes, a sensitivity analysis including only studies of ‘high quality’ was performed, with similar results obtained.
We further stratified the associations between NAFLD and an increased risk of increased CIMT by study region (figure 4A). The pooled ORs for increased CIMT were (OR: 1.63, 95% CI: 1.19 to 2.22, Pheterogeneity=0.06, I2=50%, n=7 studies) in Asian populations vs (OR: 2.70, 95% CI: 1.58 to 4.60, Pheterogeneity<0.001, I2=93.6%, n=3 studies) in Western populations (Pdifference=0.15). Likewise, the pooled Std Diff in mean CIMT were 0.75 (95% CI: 0.31 to 1.17) in Asian populations (n=12 studies) vs 0.67 (95% CI: 0.25 to 1.09) in Western populations (Pdifference=0.83) (figure 4B). Lastly, when analysing the subset of studies conducted on participants with T2DM, no Std Diff in CIMT means were found between those with and without NAFLD (Std Diff in means: 0.99, 95% CI:−0.21 to 2.20, n=7 studies) (online supplemental figure 1).
Effect of NAFLD on CAC score
Figure 5 summarises the studies investigating the associations between NAFLD and CAC score. Twenty-two studies, with a total of 136 294 individuals, assessed the effect of NAFLD on CAC score. Sixteen studies investigated the cross-sectional associations between NAFLD and the presence of CAC score >0,22 26 27 29 31 32 47 60 84–91 five studies investigated the cross-sectional associations between NAFLD and the presence of CAC score >100,60 80 85 92–94 and four studies investigated the longitudinal influence of NAFLD on CAC score progression/development.31 34 95 96
Compared with participants without NAFLD, the presence of NAFLD was significantly associated with the presence of both CAC score >0 (pooled OR: 1.21, 95% CI 1.12 to 1.32, Pheterogeneity=0.018, I2=47.7%), and CAC score >100 (pooled OR: 1.28, 95% CI 1.01 to 1.63, Pheterogeneity=0.015, I2=67.8%), (figure 5A). Likewise, NAFLD was significantly associated with the development/progression of CAC with a pooled OR of 1.26 (95% CI 1.04 to 1.52, Pheterogeneity=0.34, I2=10.6%) (figure 5B).
As with the CIMT analysis, we further stratified the associations of NAFLD with CAC score >0 based on ethnicity (figure 6). The pooled associations between NAFLD and CAC score >0 were (OR: 1.21 95% CI 1.10 to 1.33, Pheterogeneity=0.15, I2=31.7%, n=10 studies) in Asian populations vs (OR: 1.20 95% CI 1.03 to 1.38, Pheterogeneity=0.004, I2=73%, n=5 studies) in Western populations (Pdifference=0.98). There were too few studies to conclusively compare ethnic differences for the associations with CAC score >100, or for the progression/development of CAC.
Evaluation of publication bias
When assessing the studies that investigated the relationships between NAFLD and CIMT, the funnel plot showed asymmetry (online supplemental figures 2 and 3), with studies favouring increased Std Diff in means CIMT (Egger’s, p<0.05) and positive ORs for increased CIMT (Egger’s, p=0.002). For studies investigating the relationships between NAFLD and CAC outcomes (online supplemental figures 4 and 5), the funnel plots excluded bias with symmetrical distribution of studies on both sides of the mean, while the Egger’s test was non-significant (p=0.07 for CAC presence, and p=0.15 for CAC progression/development).
In this meta-analysis, we evaluated the associations of NAFLD with two established markers of subclinical atherosclerosis, synthesising the results of 64 published studies with a total of 172 385 patients. In line with existing literature, we have demonstrated that subjects with NAFLD have an increased risk of prevalent subclinical atherosclerosis than those without, even after adjustment for common cardiometabolic risk factors. Our subgroup analyses also revealed these associations to be consistent across both Western and Asian populations. This is also the first meta-analysis to demonstrate that subjects with NAFLD are at increased risk of development and progression of subclinical atherosclerosis. This may provide additional insights into screening and surveillance strategies for patients with NAFLD,2 potentially identifying higher-risk NAFLD populations, and may also provide further insight into the role of NAFLD in the development of CVD.
Our meta-analysis serves as a timely update to build on the previous work of Zhou et al, Kapuria et al and Jaruvongvanich et al,23–25 incorporating the results of over 21 new studies published from 2016 and 2020, comprising over 100 000 participants (~50 000 of which have NAFLD). The inclusion of these new studies enables us to conduct a more robust analysis of the differences between ethnic populations, with a larger number of studies conducted in both Western and Asian populations. Our overall findings of the associations between NAFLD and an increased risk of subclinical atherosclerosis (as measured by CIMT and/or CAC score) are in agreement with existing literature, further reinforcing the findings of previous studies and meta-analyses.11 12 23–25 97 In addition to these associations with subclinical atherosclerosis, other meta-analyses have also found NAFLD to be significantly associated with increased cardiovascular mortality, coronary artery disease (CAD), incident CVD events, and other subclinical manifestations of CVD including abnormalities in myocardial metabolism, ventricular structure and function.98–100 Our findings reiterate how the increased risk of CVD in patients with NAFLD can be attributed to an increased underlying subclinical atherosclerotic burden, and suggest that patients with NAFLD should be considered at high risk of atherosclerotic CVD.
Interestingly, we did not observe differential associations between NAFLD and both CAC or CIMT across Asian and Western populations. Our subgroup analyses found similar associations between NAFLD and CAC in both Asian (OR: 1.21 (1.10 to 1.33)) and Western regions (OR: 1.20 (1.03 to 1.38)), with a Pdifference=0.98. Likewise, similar associations between NAFLD and increased CIMT were found across both regions. Despite literature suggesting ethnic differences in the pathogenesis, severity and outcomes of NAFLD,14 17 101 remarkably few studies have specifically investigated these ethnic differences in the context of associations with subclinical atherosclerosis. The Multi-Ethnic Study of Atherosclerosis found a positive association between NAFLD and both CAC and increased CIMT in white and Hispanic individuals, but not in Chinese individuals.21 22 While we did not specifically look at ethnic differences, our results show that NAFLD serves as an important atherogenic risk factor in both Western and Asian populations.
The associations between NAFLD and atherosclerotic CVD were originally considered epiphenomena due to a shared confluence of metabolic risk factors.102 However, increasingly, evidence has now recognised that NAFLD is an independent risk factor for CVD, with NAFLD thought to play an active role in the systemic release of proatherogenic and proinflammatory mediators, with additional contributions to insulin resistance and abnormal atherogenic lipid profiles, all of which increase the risk of atherogenesis.3 5 8 These potential pathways and mechanisms are covered in detail in other reviews.2 3 103 Nonetheless, the interplay between NAFLD, MetS, diabetes and CVD remains complex. Evidence on the effect of NAFLD on subclinical atherosclerosis within subjects with T2DM, for example, remains equivocal.33 52 71 77 In our subgroup analysis of studies conducted within populations with T2DM, our forest plots did not show significant Std Diff in mean CIMT between those with NAFLD and those without (p=0.107). Diabetes is a potent risk factor for both CAD and CVD, and may have thus masked subtler associations between NAFLD and subclinical atherosclerosis. Alternatively, this may also highlight the role of insulin resistance in mediating the relationship between NAFLD and atherosclerosis.104
Only recently have studies begun to investigate the longitudinal associations between NAFLD and CAC progression/development, with this paper being the first meta-analysis to synthesise the results of four studies published from 2016 onwards.31 34 95 96 We demonstrated that patients with NAFLD are at greater risk of development/progression of CAC, even after adjustment for known confounders. While our results do not elucidate the exact pathophysiological mechanisms by which NAFLD may affect CAC development/progression, they do provide insight into the causal relationship between NAFLD and subclinical atherosclerosis. It should be noted that Park et al found differential associations between NAFLD and CAC development and progression, reporting that NAFLD might play a role in the early development of atherosclerosis, but not in the progression to more severe degrees of atherosclerosis.34 Future studies may be warranted to confirm such observations.
Strengths of our study include the large participant numbers, the assessment of various modalities of subclinical atherosclerosis including CIMT and CAC, the large number of studies from both Western and Asian populations enabling robust analysis of regional differences, and our analysis of not just cross-sectional, but longitudinal outcomes (CAC development and progression). Nonetheless, our results should be interpreted with caution, taking into consideration certain limitations. Heterogeneity was consistently present across the different subclinical atherosclerotic outcomes. This can be attributed to differences in study design, population characteristics, the use of different cut-off definitions for both CAC and increased CIMT, the adjustment for different cardiometabolic confounders and the different modalities of NAFLD diagnosis. In addition, even though liver biopsy remains the gold standard for NAFLD evaluation, US was the most common modality used in the NAFLD assessment in the included studies, and is cited to have diminished accuracy when it comes to the diagnosis of milder hepatic steatosis.105 106 While we did not find regional differences in the results, we could not perform actual ethnic comparisons as these data were not available. Whether these regional data accurately reflect ethnic data is uncertain and also the influence of cultural and socioeconomic factors cannot be quantified. Nevertheless, this provides one of the first combined regional comparison of such results. Finally, potential publication bias exists with regard to the studies investigating CIMT-related outcomes.
In conclusion, this meta-analysis reports a significant positive association between NAFLD and subclinical atherosclerosis, as defined by increased CIMT and CAC scores. These observed associations are not just cross-sectional, but also longitudinal, and are seen across both Western and Asian populations. These results re-emphasise the importance of early risk evaluation and prophylactic intervention measures to preclude progression to clinical CVD in NAFLD.
Data availability statement
Data are available upon reasonable request.
Patient consent for publication
Ethical approval is not applicable for this systematic review/meta-analysis.
MYZW and JJLY contributed equally.
Contributors MYZW—conception of idea, crafting of research question, design of inclusion/exclusion criteria, collection of data (literature search), statistical analysis, quality (risk of bias) evaluation, figure creation, writing of the manuscript, and writing and drafting of the manuscript, guarantor. JJLY—conception of idea, crafting of research question, design of inclusion/exclusion criteria, collection of data (literature search), drafting of the manuscript and editing of the manuscript. RS—design of inclusion/exclusion criteria, collection of data (literature search), statistical analysis, quality (risk of bias) evaluation, figure creation and editing of the manuscript, guarantor. MC—conception of idea, crafting of research question, design of inclusion/exclusion criteria and editing of the manuscript. GBBG—conception of idea, crafting of research question, design of inclusion/exclusion criteria and editing of the manuscript. KKY—conception of idea, crafting of research question, design of inclusion/exclusion criteria, editing of the manuscript and guarantor.
Funding KKY has received research funding from Medtronic, Boston Scientific, Amgen, AstraZeneca, Shockwave Medical (all significant, via institution); consulting or honoraria fees (all modest) from Medtronic, Boston Scientific, Abbott Vascular, Amgen, Bayer, Novartis; speaker or proctor fees from Abbott Vascular, Boston Scientific, Medtronic, Philips, Shockwave Medical, Alvimedica, Menarini, AstraZeneca, Amgen, Bayer.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
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.