Background Fabry Disease (FD), an X linked lysosomal storage disease due to pathogenic α-galactosidase A (GLA) mutations, results in two major subtypes, the early-onset Type 1 ‘Classic’ and the Type 2 ‘Later-Onset’ phenotypes. To identify previously unrecognised patients, investigators screened cardiac, renal and stroke clinics by enzyme assays. However, some screening studies did not perform confirmatory GLA mutation analyses, and many included recently recognised ‘benign/likely-benign’ variants, thereby inflating prevalence estimates.
Methods Online databases were searched for all FD screening studies in high-risk clinics (1995–2017). Studies reporting GLA mutations were re-analysed for pathogenic mutations, sex and phenotype. Phenotype-specific and sex-specific prevalence rates were determined.
Results Of 67 studies, 63 that screened 51363patients (33943M and 17420F) and provided GLA mutations were reanalysed for disease-causing mutations. Of reported GLA mutations, benign variants occurred in 47.9% of males and 74.1% of females. The following were the revised prevalence estimates: among 36820 (23954M and 12866F) haemodialysis screenees, 0.21% males and 0.15% females; among 3074 (2031M and 1043F) renal transplant screenees, 0.25% males and no females; among 5491 (4054M and 1437F) cardiac screenees, 0.94% males and 0.90% females; and among 5978 (3904M and 2074F) stroke screenees, 0.13% males and 0.14% females. Among male and female screenees with pathogenic mutations, the type 1 Classic phenotype was predominant (~60%), except more male cardiac patients (75%) had type 2 Later-Onset phenotype.
Conclusions Compared with previous findings, reanalysis of 63 studies increased the screenee numbers (~3.4-fold), eliminated 20 benign/likely benign variants, and provided more accurate sex-specific and phenotype-specific prevalence estimates, ranging from ~0.13% of stroke to ~0.9% of cardiac male or female screenees.
- fabry disease
Statistics from Altmetric.com
Fabry disease (FD) is an X linked lysosomal storage disease caused by pathogenic mutations in the α-galactosidase A gene (GLA) that result in the absent or markedly deficient activity of its encoded lysosomal hydrolase, α-galactosidase A (α-GalA) (MIM #301500).1 2 The enzyme deficiency leads to the progressive accumulation of glycosphingolipids with terminal α-galactosyl moieties, primarily globotriaosylceramide (Gb3) and its deacylated derivative, lyso-Gb3,3 in fluids and tissue lysosomes throughout the body.1
Clinically, there are two major subtypes: the early-onset Type 1 ‘Classic’ and Type 2 ‘Later-Onset’ phenotypes.4 Affected males with the Type 1 Classic phenotype have little or no functional α-GalA enzymatic activity, marked microvascular endothelial glycosphingolipid accumulation, and childhood or adolescence onset of clinical manifestations including acroparesthesias, angiokeratomas, hypohidrosis, gastrointestinal symptoms and a characteristic corneal dystrophy.1 2 With advancing age, the progressive accumulation of the glycosphingolipid substrates, especially in the microvascular endothelial cells, cardiomyocytes and podocytes, leads to cardiac, renal and/or cerebrovascular disease and early demise. Prior to renal replacement, enzyme replacement or pharmacological chaperone therapies, the mean age of death of classically affected males was ~40 years.5
In contrast, affected males with the Type 2 Later-Onset phenotype have residual α-GalA activity, little, if any, microvascular endothelial glycosphingolipid accumulation, and therefore lack the early manifestations of males with the Type 1 Classic phenotype.6–8 They progressively accumulate the glycosphingolipid substrates, especially in cardiomyocytes and podocytes, and typically develop renal and/or cardiac disease in their fourth to seventh decades of life.1 2 6 8 Notably, the Type 2 Later-Onset males are often misdiagnosed and/or are unrecognised throughout life. Most index cases have been identified as having FD by α-GalA enzyme screening in cardiac, haemodialysis and renal transplant clinics, followed by family studies.
In 2010, Linthorst and colleagues9 reviewed the results of 20 FD screening studies reported from 1995 to 2008 to provide prevalence estimates of previously unrecognised patients with FD in clinic populations. These studies included patients tested for α-GalA deficiency using dried blood spots (DBS), plasma and/or leucocyte enzyme assays alone, or confirmed by GLA mutation analyses (see online supplementary table 1). Linthorst and Ginsberg10 updated their previous conclusions citing two additional studies that excluded patients with GLA polymorphisms (eg, the variant encoding p.D313Y) and intronic mutations, suggesting that such mutations were not pathogenic and inclusion inflated the previously calculated prevalence estimates for FD, particularly in stroke cohorts. Of note, certain screening studies in the original Linthorst report and in more recent studies were prescreened to identify patients more likely to have FD (eg, Rolfs et al,11 Havndrup et al12 and Coutinho et al13) as preselected studies would have higher prevalence values. Screening studies that did not verify deficient enzyme activities with confirmatory mutation analyses also could increase prevalence values, particularly since certain GLA mutations result in low α-GalA activity (ie, pseudodeficiency alleles). In retrospect, many studies overstated the prevalence estimates since they included GLA mutations that are now recognised as relatively common variants that are benign or likely benign (see online supplementary table 2, eg, variants encoding p.E66Q, p.R118C, p.S126G, p.A143T and p.D313Y).14–26
Supplementary file 1
Supplementary file 2
Here, we reanalyse the screening reports for FD in haemodialysis, renal transplantation, cardiac and stroke clinics from 1995 through September 2017. Of the 67 reports, 63 confirmed the screenees with low α-GalA activities followed by GLA sequencing analyses. The estimated prevalence values in each report were reanalysed after separating pathogenic mutations from those that are now reclassified as benign or likely benign. Here, we report the overall estimated prevalence values of the patients screened in the four clinic populations with and without the benign and likely benign variants. Also, we classified the phenotype or likely phenotype of each mutation to determine the percentages of patients in each clinic population having the type 1 classic or type 2 later-onset phenotypes. These updated sex-specific and phenotype-specific data provide more accurate estimates of the prevalence of FD-affected males and heterozygotes in these high-risk populations.
Identification of FD screening publications
Online databases, including PubMed and Medline, were searched using the keywords ‘Fabry AND screening’, ‘Fabry AND cardiac’, ‘Fabry AND dialysis’, ‘Fabry AND renal transplant’, and Fabry AND ‘stroke’. Additional screening studies were identified from other published screening reports. Not included were single-mutation family studies, small case cohorts and screening studies that did not report specific GLA mutations.
Data collection and analysis
For each reported study, the following information was collected: study’s first author, date of publication and country; clinic population studied; number of males and females screened; patient preselection criteria, if any; methods of initial identification and confirmation; methods for α-GalA enzyme assay and/or GLA mutation determinations; GLA mutations identified; and number of patients identified with GLA pathogenic versus benign/likely benign mutations. For each published study, mutations were separated by sex and assigned to the Type 1 or 2 phenotypes. Percentages of FD males and heterozygotes with pathogenic GLA mutations, as well as their likely phenotypes, were determined for each clinic population.
Determination of FD phenotype
Assignment of the phenotype for each GLA pathogenic mutation was based on peer-reviewed publications as listed in the Human Gene Mutation Database (https://portal.biobase-international.com/hgmd/pro/all.php),27 mutation-specific searches of PubMed, review of published clinical information, biochemical findings, in vitro α-GalA expression, structural analyses, and/or relevant clinical and biochemical information from patients seen for over 40 years at the Mount Sinai International Center for Fabry Disease or tested at the Mount Sinai Genetic Testing Laboratory. Each mutation was assigned as Type 1 Classic, Type 2Later-Onset or benign based on available clinical, biochemical and pathological studies (see the International Fabry Disease Genotype/Phenotype Database (dbFGP.org)).
In vitro α-GalA enzyme activity determinations
The full-length wild-type (WT) α-GalA cDNA, including the 5' untranslated region (UTR), was cloned into the expression vector pCDNA3 for which the neomycin resistance gene has been substituted by the gene coding for the enhanced Green Fluorescent Protein (eGFP), allowing its expression under the control of the SV40 promoter. Individual constructs carrying the mutation encoding p.N53K, p.E66Q, p.R118C, p.S126C, p.A143T, p.Y152H, p.F229V, p.M290I (c.870G>A), p.M290I (p.870G>C), p.D313Y, p.Q330R, p.R356Q and p.A368T were generated by site-directed mutagenesis (Stratagene, La Jolla, California, USA) using sequence-specific primers. All constructs were confirmed by sequencing. The GLA constructs (1 mg) were transfected using the FuGENE HD Transfection Reagent (Roche Molecular Biochemicals, Indianapolis, Indiana, USA) according to the manufacturer’s protocol, into COS-7 cells (ATCC), which were maintained in Dulbecco’s Modified Eagle Medium (Invitrogen, Carlsbad, California, USA) supplemented with 10% fetal bovine serum at 37°C with 5% CO2. COS-7 cells were harvested 48 hours after transfection. The α-GalA enzyme activity was determined as previously described28 with the following modifications. Enzymatic activities were first normalised against the protein concentration in the lysates, which was determined with the Bradford Protein Assay (Bio-Rad, Hercules, California, USA). The endogenous α-GalA activity, as measured in empty vector-transfected cell lysates, was subtracted from the activity in mutant or WT-transfected cell lysates. Transfection efficiency was measured by the amount of Green Fluorescent Protein (GFP) fluorescence in the cell lysates quantitated using the Modulus fluorometer (Turner BioSystems, Sunnyvale, California, USA), and the activity values were corrected accordingly. All assays were done in at least triplicates. The values obtained for each mutant were calculated as per cent of the WT activity.
From 1 January 1995 through 30 September 2017, a total of 67 publications reporting FD screening results in disease clinics were identified. Four studies that did not report specific GLA mutations were excluded.11 29–31 Of the remaining 63 studies, 27 reports screened 36 820 (23954 males and 12866 females) haemodialysis patients,8 13 32–56 3 studies screened 3074 (2031 males and 1043 females) renal transplant patients,57–59 17 reports screened 5491 (4054 males and 1437 females) patients with left ventricular hypertrophy (LVH) and/or hypertrophic cardiomyopathy (HCM),7 12 25 60–73 and 16 studies screened 5978 (3904 males and 2074 females) patients with ischaemic or cryptogenic strokes.26 74–88
The detailed results from each report in these clinic populations are summarised in online supplementary table 1. The overall frequencies of affected males and heterozygous females with pathogenic GLA mutations for the haemodialysis, renal transplant, cardiac and stroke patients screened are shown in tables 1 and 2, respectively. These tables also compare the findings of Linthorst et al9 with the updated findings including the total number of male and female patients screened for each specialty clinic cohort, the original number of GLA mutations identified, and the reanalysed number and per cent of pathogenic mutations and benign/likely benign variants. Online supplementary tables 3 and 4 identify the benign/likely benign variants for males and females, respectively.
Prevalence among renal haemodialysis and transplant screenees
Among the 23954 males screened at haemodialysis clinics, there were 101 with GLA mutations (0.42% of screenees). Among these 101 males, 51 (50.5%) had benign variants, including variants encoding p.E66Q (17, all Asian), p.R118C (9), c.870G>C encoding p.M290I (8), p.A368T (7), p.A143T (2), p.D313Y (2) and c.870G>A encoding p.M290I (2), and IVS6-22C>T (1), the regulatory variant c.1–10C>T (1), and variants encoding p.F229V (1) and p.R356Q (1) (see online supplementary tables 1 and 3). When only the 50 pathogenic GLA mutations were included, the prevalence of affected males decreased to 0.21% of screenees, of whom 66% and 34% had the Type 1 Classic and Type 2 Later-Onset phenotypes, respectively.
Similarly, of the 2031 renal transplant males screened, only 11 screenees (0.54%) were reported to have GLA mutations. Of these, five (0.24%) had pathogenic mutations, with three and two having the Type 1 Classic and Type 2 Later-Onset phenotypes, respectively, while six had benign/likely benign variants encoding p.A143T (1), and p.Q330R (1) and 5’−10C>T (4).
Since most of the female haemodialysis and transplant patients were initially screened by enzyme assays, only those with low activity were detected (the per cent of mean normal cut-off values for inclusion varied by study). Of the 12866 female haemodialysis screenees, GLA mutations were identified in 87 (0.68%), of whom only 19 (0.15%) had pathogenic mutations, with 13 (68.4%) and 6 (31.6%) having the Type 1 Classic and Type 2 Later-Onset phenotypes, respectively. Benign variants accounted for 78.3% of the originally reported mutations among female haemodialysis screenees. Of the 1043 female transplant patients screened, none had a pathogenic mutation, while three had benign GLA variants, encoding p.S126G (1) and p.D313Y (2) (see online supplementary tables 1 and 4).
Prevalence among cardiac screenees
Among the 4054 males screened with LVH or HCM, there were 49 (1.2%) with GLA mutations. Of these mutations, 11 were benign or likely benign, including variants encoding p.E66Q (3), p.A143T (2), Y152H (1) and p.D313Y (5) (see online supplementary tables 1 and 3). Among the 38 (0.94%) males with pathogenic GLA mutations, 29 (76.3%) had the Type 2 phenotype.
Among the 1437 female cardiac patients screened, 22 (1.53%) were reported to be GLA mutation-positive. Among these patients, nine (40.9%) had benign/likely benign variants, including variants encoding p.R118C (1), p.A143T (5), p.D313Y (2), and the intron 4 splice-site variant, c.639+6A>C (1). When limited to pathogenic mutations, there were 13 females (0.90%) of whom 8 (61.5%) had mutations resulting in the Type 1 Classic phenotype, consistent with the identification of Type 1 heterozygotes with lower GLA activities and more severe disease.
Prevalence among stroke screenees
In the 16 reviewed studies of patients with primarily cryptogenic or ischaemic strokes that had confirmatory GLA mutation analyses, 3904 males were screened and 26 (0.67%) were reported to have GLA mutations. Of these, only five males (0.13%) had pathogenic mutations, three causing theType 1 phenotype and two having the Type 2 phenotype. Notably, 21 (80.8%) of the 26 reported GLA mutations were benign or likely benign variants: encoding 5’-untranslated region of exon 1, g.1136C>T (5’−44C>T) (1), p.E66Q (10, all Asians), p.R118C (4), p.A143T (a male with normal leucocyte α-GalA activity) and p.D313Y (5).
Similarly, 2074 females were screened, with 23 (1.11%) reported to have GLA mutations, of whom only 3 were pathogenic, specifically mutations encoding p.M1T, p.L415R and c.548-3del9, all three resulting in the Type 1 phenotype. Twenty of the GLA mutation-positive female stroke screenees had benign or likely benign variants encoding: variant 5’−44C>T (1), p.S126G (1), p.E66Q (2, both of Asian descent), p.A143T (2), p.R118C (3) and p.D313Y (11).
From January 1995 through September 2017, there have been 67 reports describing studies that screened patients in haemodialysis, renal transplant, cardiac and stroke clinics for previously unrecognised affected males and heterozygotes with FD. Nakao et al7 were the first to demonstrate that FD males with LVH were readily identified by their markedly deficient plasma α-GalA activities, with confirmation by GLA gene sequencing or transcript studies. Subsequently, Utsumi et al32 in 2000 extended this approach to haemodialysis patients. The development of the DBS assay for α-GalA activity89 further facilitated screening studies due to the ease of collection, shipment and assay. Such screening studies intensified when safe and effective enzyme replacement therapy was approved in Europe in 2001, in the USA in 2003, and in Korea and Japan in 2004.
In 2010, Linthorst and colleagues9 reviewed 20 screening studies and determined the prevalence of unrecognised patients with FD in renal, cardiac and stroke clinics (tables 1 and 2 and online supplementary table 1). However, a major limitation of these and more recent screening studies was that the pathogenicity of the almost 1000 reported GLA mutations27 had not been established, and currently designated benign or likely benign variants were considered pathogenic, including c.196G>C (encoding p.E66Q) in Japanese and Koreans, c.352C>T (p.R118C) in Hispanics and Caucasians, and c.376A>G (p.S126G), c.427G>A (p.A143T) and c.937G>T (p.D313Y) in Caucasians. Recent studies have clarified that these and other variants are benign or likely benign by their activities expressed in vitro,16 17 24 28 demonstration of normal Gb3 or Lyso-Gb3 concentrations in tissues and/or fluids,14 15 23 90 91 lack of histological and/or ultrastructural evidence of glycosphingolipid accumulation in biopsied tissues of affected males,14 19 22 23 25 and by deep clinical phenotyping.92 In addition, the recent availability of large exonic and genomic databases (ExAC, http://exac.broadinstitute.org; and GnomAD, http://gnomad.broadinstitute.org)93 provided the allele frequencies in individuals from different ethnic, racial and demographic populations, indicating that many variants were rare but relatively common (0.3%–0.02%).
In contrast to the total of 15089 (9973 males and 5116 females) screenees in the 20 studies reported in 2010,9 the total number of screenees analysed in the 63 reports with GLA confirmation (51363 (33943 males and 17420 females)) was ~3.4 fold greater overall and in both genders. The numbers of males screened in haemodialysis, renal transplant, cardiac and stroke clinics were 3.3-fold, 1.3-fold, 5.7-fold and 7.9-fold greater, respectively. Similarly, the numbers of female screenees in the above four clinic cohorts increased 3.1-fold, 2.6-fold, 6.7-fold and 6.3-fold, respectively. Thus, these additional studies markedly increased the available data to more accurately estimate the prevalence rates in these clinic populations.
By eliminating the benign and likely benign variants from each of the 63 studies that reported GLA mutations, the previously published prevalence estimates were decreased in all males and heterozygotes in clinic populations except for female haemodialysis patients. Of the 51363 patients screened, the greatest number was from haemodialysis clinics (~36000, ~65% males), of which the prevalence of males with pathogenic GLA mutations was reduced to 0.21% (1 in 490) from the original estimated prevalence of 0.33% (1 in 300).9 About 65% of males with pathogenic lesions had the Type 1 Classic phenotype. Among the >12 000 female haemodialysis patients screened, the previous prevalence of 0.10% (1 in 1000)9 increased slightly to 0.15% (1 in 670), of which ~68% had theType 1 Classic phenotype. Among the renal transplant patients, no females were found to have a pathogenic GLA mutation, while the prevalence among transplanted males was reduced slightly from 0.38% (1 in 260)9 to 0.25% (1 in 400), with 60% havingType 1 Classic phenotype. Among over 4000 male cardiac patients, the originally published prevalence was reduced from 2.67% (1 in 37)9 to 0.94% (1 in 106), with ~75% having the Type 2 Later-Onset phenotype. Similarly, among over 1400 female cardiac patients, the original published prevalence was markedly reduced from 2.8% (1 in 35)9 to 0.90% (1 in 111). Contrary to the male cardiac cohort, however, twice as many females had the Type 1 Classic phenotype than the Type 2 Later-Onset phenotype. In the 2010 prevalence estimate for stroke,9 there were two studies that, when combined, screened only 496 males and 328 females.11 74 However, the larger study11 did not report GLA mutations, and no mutations were detected in the other study.74 Although the GLA mutations were not reported, the larger study reported that 4.23% of males and 2.13% of females with cryptogenic strokes had unrecognised FD.11 Subsequently, 15 studies reported almost 4000 additional male patients with stroke, thus markedly reducing the original 2010 male prevalence of 4.23% (1 in 24)9 to 0.13% (1 in 780), with the Type 1 Classic phenotype more frequent than the Type 2 Later-Onset phenotype. Among over 2000 female patients with stroke screened in the 14 studies reporting GLA mutations, the prevalence estimate was reduced from the 2010 rate of 2.13% (1 in 47)9 to 0.14% (1 in 690). All three female patients with pathogenic GLA mutations had the Type 1 phenotype.
It is surprising that the majority of patients identified in haemodialysis and stroke clinics had the Type 1 Classic phenotype, particularly since most of these patients should have been symptomatic as children and should have been diagnosed. Clearly paediatricians and family physicians should become familiar with the early-onset manifestations of FD, especially acroparesthesias, angiokeratoma, hypohidrosis and corneal dystrophy, which are the clinical hallmarks of the early-onset disease. Since these early-onset manifestations are typically absent in the Later-Onset Type 2 phenotype, it would be expected that the late manifestations of renal insufficiency and failure, LVH leading to HCM, and transient ischaemic attacks and strokes could go unrecognised in haemodialysis, cardiac and stroke clinics.
Among male haemodialysis patients, the prevalence of pathogenic GLA mutations was similar in Europeans (mostly Caucasians) and Asians (0.24% and 0.16%, respectively). However, the ratios of Type 1 Classic to Type 2 Later-Onset phenotypes were 2.5:1 among European Caucasians versus 1:1 among Asian FDmales. Among female haemodialysis patients, the prevalence was 11 times greater in European Caucasians (0.22%, 1 in 454) than in Asians (0.02%, 1 in 5000). While the ratio of Type 1 Cassic to Type 2 Later-Onset phenotypes was similar to European males (2.6:1), there was only one Asian female haemodialysis patient identified who had a Type 2 Later-Onset GLA mutation. Among male cardiac patients, the prevalence of pathogenic GLA mutations was similar in Caucasian and Asian (1.00% and 0.79%, respectively). In contrast to male haemodialysis patients, there were more Type 2 Later-Onset patients in both European Caucasian (twofold) and Asian (no Type 1 Classic) cardiac cohorts. Moreover, both European Caucasian and Asian male cardiac cohorts had the lowest percentage of benign mutations (~22% and 25%, respectively), compared with other clinical cohorts (ranging from 41% to 100%). Among males in the stroke cohort, 0.18% (1 in 555) of Caucasians were found to have pathogenic GLA mutations, with a 3:2 Type 1:Type 2 ratio. Among Asian male patients with stroke, all previously reported mutations were benign. Among female patients with stroke, the prevalence of pathogenic GLA mutations was 0.11% (1 in 910) in Caucasians and all mutations were Type 1. Among three reported Asian female patients with stroke, only one patient had a Type 1 pathogenic mutation. Two other patients had benign variants. There were no screening studies of Asian renal transplant patients.
In all clinic groups, the initial screening data were more reliable for males with this X linked disease, since enzyme screening, particularly with DBS assays, was clearly less reliable in females, as up to 40% of female heterozygotes have normal or near normal enzyme values.9 94 In addition, the prevalence estimates for individual clinic reports varied due to the fact that certain studies employed preselection procedures, which introduced selection bias. For example, Coutinho et al13 prescreened over 25000 haemodialysis patients with a questionnaire, and subsequently identified 2956 FD-suspected patients who were then screened by enzyme assay. The selection process also varied among patients in stroke clinics, including preselection for young or all patients with cryptogenic strokes,74 76–79 81 85 86 88 transient ischaemic attacks and ischaemic strokes,81 82 and ischaemic strokes alone.26 75 80 83 84 86–88 Notably, 11 cardiac studies selected patients with HCM, while 6 other included only patients with LVH, and 1 study72 preselected patients with HCM who had septal myectomies whose cardiac biopsies were suggestive of a lysosomal storage disorder (see online supplementary table 1). Such selection bias among the cardiac studies may in part account for the threefold to sevenfold higher FD prevalence in males and females screened than the lower rates in haemodialysis, transplant and stroke clinics (tables 1 and 2). The higher prevalence of the Type 2 phenotype among male cardiac screenees may also be due to the frequency of certain relatively frequent GLA mutations that cause the later-onset cardiac phenotype, such as c.644A>G; p.N215S.95
In conclusion, this reanalysis of all reported screening studies for FD from 1995 through September 2017 provides a more valid estimate of the prevalence of FD in high-risk populations. Benign variants were found in all cohorts, accounting for inflation of the previous prevalence values from 1.6-fold in male renal transplant patients to 33-fold and 15-fold in male and female patients with stroke. Overall, the recalculated prevalence estimates of previously unrecognised FD in these high-risk cohorts ranged from 0% in female renal transplant screenees to about 0.9% among both male and female cardiac cohorts. Of clinical importance, continued screening for FD is indicated in these high-risk populations. However, as this study indicates, screening should include GLA genotyping with phenotype determination for all patients found to have a GLA mutation to differentiate patients with the Type 1 and Type 2 phenotypes from benign variants. Patients and their at-risk family members should be screened for this X linked disease, and affected individuals should be counselled and be evaluated by FD specialists for phenotype-specific medical management and treatment.
Supplementary file 3
Supplementary file 4
Contributors DD and RJD conceived the study. DD, RS, SP and BC collected the data. DD, RS, BC and RJD analysed the data. DD, BC, MY and RJD wrote the manuscript, which was reviewed and approved by all authors.
Funding This research was supported fully from research funds from the Department of Genetic and Genomic Sciences, Icahn School of Medicine at Mount Sinai.
Competing interests RJD is a consultant for Amicus Therapeutics, Sanofi Genzyme and Sangamo Therapeutics, and receives a training grant and royalties from Sanofi Genzyme and Shire.
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.