This article was corrected | View correction
The disease mechanism of bipolar disorder remains unknown. Recent studies
have provided evidence for abnormal gene expression in bipolar disorder.
To determine the expression of 12 558 nuclear genes in the human
hippocampus in healthy control subjects and those with bipolar disorder or
We used gene arrays to study messenger RNA expression. Data were verified
with a real-time quantitative polymerase chain reaction assay.
We studied 10 healthy control subjects, 9 subjects with bipolar disorder,
and 8 subjects with schizophrenia.
The expression of nuclear messenger RNA coding for mitochondrial proteins
was significantly decreased in the hippocampus in subjects with bipolar disorder
but not in those with schizophrenia. Subjects with bipolar disorder were characterized
by a pronounced and extensive decrease in the expression of genes regulating
oxidative phosphorylation and the adenosine triphosphate–dependent process
of proteasome degradation.
These findings point toward a widespread dysregulation of mitochondrial
energy metabolism and downstream deficits of adenosine triphosphate–dependent
processes in bipolar disorder.
Bipolar disorder affects approximately 0.5% of the world population,often leading to recurrent illness and a marked decline in social function.1 The clinical features of bipolar disorder (ie, recurrentepisodes of depression and either full-blown mania with frank psychosis ormilder bouts of hypomania) have long been recognized.2 However,the etiologic and disease mechanisms remain unknown. For example, bipolardisorder shows a high degree of heritability (approximately 0.8%), and severalstudies have reported linkage of bipolar disorder to chromosomal loci, butnot a single locus has repeatedly been linked to bipolar disorder.3
Recent spectroscopic studies have provided evidence for bipolar disorderas a disease of mitochondrial energy metabolism,4 includingdecreased pH5 and decreased high-energy phosphates6- 7 in the frontal and temporal lobes ofthese patients. Such mitochondrial dysfunction in bipolar disorder could bedue to an abnormal expression of nuclear or mitochondrial genes coding formitochondrial proteins.8 In this article, wereport that the expression of nuclear messenger RNA (mRNA) coding for mitochondrialproteins is significantly decreased in the hippocampus in bipolar disorderbut not in schizophrenia.
We analyzed the expression of 12 558 nuclear genes in 3 study groups:healthy controls, subjects with bipolar disorder, and subjects with schizophrenia.Brain specimens were obtained from the Harvard Brain Tissue Resource Center(McLean Hospital, Belmont, Mass) and initially consisted of 10 subjects ineach diagnostic group. Each control subject was matched with 1 subject whohad schizophrenia and 1 who had bipolar disorder for age and postmortem intervalto ensure homogeneity of the groups. One subject with bipolar disorder and1 subject with schizophrenia were excluded from the study because they didnot provide sufficient RNA quality, as assessed by the 3′/5′ ratioof glyceraldehyde-3-phosphate dehydrogenase (>4), 3′/5′ ratioof β-actin (>4), and percentage (<37%) of "gene-present calls" (thepercentage of genes on the array that were above the detection limit in asample).
All diagnoses were established by 2 psychiatrists at the Harvard BrainTissue Resource Center via retrospective review of all available medical recordsand extensive questionnaires about social and medical history completed byfamily members of the donors. We applied the criteria of Feighner et al9 for the diagnosis of schizophrenia and that of the DSM-III10 for the diagnosisof schizoaffective disorder and bipolar disorder. Probands with a documentedhistory of substance dependence or neurological illness were excluded fromthe study. During our study it became evident that the documentation of 1subject with schizophrenia was not sufficient to verify the diagnosis, andthat case had to be excluded.
One hemisphere of each brain underwent a comprehensive neuropathologicexamination, which revealed no evidence of stroke, tumor, infection, or neurodegenerativechanges. After exclusion of 3 cases (insufficient RNA quality in 2 cases andinsufficient documentation of the psychiatric history in 1 case), the finalsample sizes were 10 control subjects, 9 subjects with bipolar disorder, and8 subjects with schizophrenia (Table 1).
All brains were transported on wet ice and dissected immediately onarrival by specially trained staff at the Harvard Brain Tissue Resource Centerusing a standard protocol (see http://www.brainbank.mclean.org/ fordetails). A coronal block of the hippocampus was obtained at the level ofthe lateral geniculate nucleus, frozen in liquid nitrogen vapor, and storedat −80°C. Mean ± SD storage time was 31 ± 14 monthswith no significant difference between groups. Blocks were trimmed to includeonly the dentate gyrus and cornu ammonis sectors 1 through 4 without adjacentwhite matter of the parahippocampal gyrus. Twenty-five slices (10 µmthick) were cut from each hippocampal block in a cryostat and used for RNAextraction.
Human hippocampal RNA was prepared according to the protocol providedby Affymetrix (Santa Clara, Calif). The RNA was extracted from 50 to 100 mgof tissue with an extraction kit (RNAgents kit; Promega, Madison, Wis). Thetotal yield of RNA was the same in all 3 groups. The RNA quality was assessedusing an analytical gel and a bioanalyzer (Agilent Technologies, Palo Alto,Calif). We used 8 µg of total RNA for complementary DNA synthesis witha double-stranded complementary DNA synthesis kit (SuperScript; InvitrogenCorp, Carlsbad, Calif), and in vitro transcription was performed with an RNAtranscript labeling kit (Enzo IVT kit; Enzo Biochem, Farmingdale, NY). Boththe schizophrenia group and the bipolar disorder group had a 20% lower yieldof biotinylated RNA. Whereas the difference was not significant in the bipolardisorder group, it reached significance in the schizophrenia group. BiotinylatedRNA was fragmented and hybridized to the HG-U95Av2 array (Affymetrix) overnightat 45°C and stained on the washing station with streptavidin-phycoerythrin(Molecular Probes, Eugene, Ore) followed with a biotinylated antistreptavidinantibody (Vector Laboratories, Burlingame, Calif) and a second round of streptavidin-phycoerythrin.
Tissue preparation and RNA extraction were performed in a single batchby the same investigator to limit experimental variability. The order of sampleswas randomized, investigators were blinded to diagnoses, and the sample codewas broken before the arrays were loaded onto the washing station to enablethe investigator to randomize samples on the washing station modules.
Samples were analyzed in diagnostic groups using the dChip program (http://www.dchip.org).11 Model-basedexpression was performed on perfect match–only data. A control samplewith average intensity was chosen for normalization. We found no significantdifferences in the quality control criteria provided by the Data Mining Tool(Affymetrix) and dChip analyses (3′/5′ ratios for glyceraldehyde-3-phosphatedehydrogenase and β-actin as well as scaling factor and background) orin the ratio of 28S/18S ribosomal RNA obtained with the bioanalyzer. A significantdifference was found in gene-present calls, which were lower (P = .04) in the bipolar disorder group (Table 1).
We explored expression profiles revealed by the dChip analysis furtherwith the GenMAPP and MAPPfinder (http://www.genmapp.org) programs.GenMAPP was used to draw maps of genes in functionally related groups.12 The MAPPfinder program was used to find regulationtrends in groups of genes organized according to biological process, molecularfunction, or cellular component, as defined by the Gene Ontology Consortium(http://www.geneontology.org). The following criteria were chosenfor the MAPPfinder analysis: P<.02, with gene-presentcalls in more than 50% of the samples and at a fold induction higher than1.1. For result verification, data were also computed with Affymetrix DataMining Tool software version 3.0.
For real-time quantitative polymerase chain reaction (PCR), complementaryDNA was synthesized from 1 µg of total RNA with a synthesis system (SuperScriptFirst-Strand Synthesis System for real-time quantitative PCR; Invitrogen Corp)using oligonucleotide deoxythymidine as the primer. A primer set for eachgene was designed with the help of Primer3 software (http://www.genome.wi.mit.edu/cgi-bin/primer/primer3.cgi). Amplicons were designed to be between 100 and 150 base pairs in length.Melt curve analysis and polyacrylamide gel electrophoresis were used to confirmthe specificity of each primer pair. The real-time quantitative PCR reactionwas performed in accordance with described procedures13 (DNAEngine Opticon; Opticon Monitor Data Analysis Software version 1.4; MJ Research,Waltham, Mass) with a PCR kit (DyNAmo SYBR Green real-time quantitative PCRkit; Finnzymes, Espoo, Finland) in a volume of 25 µL, with 2.5 µLof 1:5 diluted complementary DNA samples and 0.3-µM primers. The PCRcycling conditions were initially 95°C for 10 minutes followed by 49 cyclesat 94°C for 30 seconds, 55°C for 30 seconds, and 72°C for 30 seconds.Data were collected between 72°C and 79°C depending on amplicon meltingtemperature. A melt curve analysis was performed at the end of each real-timequantitative PCR experiment. Dilution curves were generated for each primerin every experiment by diluting complementary DNA twice from a control samplewith a ratio of 1:3, yielding a dilution series of 1.00, 0.33, and 0.11. Thelogarithm of the dilution value was plotted against the cycle threshold value.Blanks were run with each dilution curve to control for cross-contamination.Dilution curves, blanks, and samples were run in duplicate. Reported valueswere normalized to the internal control human filamin A α (accessionnumber NM_001456), an actin-binding protein. Human filamin A α was notregulated in the gene array or quantitative PCR analysis. Seven control samplesand 6 bipolar disorder samples available from the original group were usedfor real-time quantitative PCR.
The identical real-time quantitative PCR parameters were used for ananalysis of 16 frontal lobe specimens (8 control subjects and 8 with bipolardisorder from the original study sample). Cortical tissue was removed fromBrodmann area 9, and RNA was extracted as detailed previously.
We initially limited our analysis to genes expressed in at least 60%of all cases, with at least a 1.2-fold differential expression at a 90% confidencelimit and significance level of P<.01. These statisticalthresholds were exceptionally stringent (false discovery rate, 2.9%) and werenot met by a single gene in the schizophrenia group. In contrast, the expressionof 43 genes was decreased in bipolar disorder (Table 2). Using more liberal statistical thresholds, we found evidencefor increased and decreased gene expression in both groups. However, for thepurpose of this article we will focus on these 43 genes, in which we discovereda striking pattern: 18 genes (42%) coded for mitochondrial proteins. Theseincluded subunits of complexes I (nicotinamide adenine dinucleotide dehydrogenasein 1 gene), IV (cytochrome-c oxidase in 1 gene),and V (adenosine triphosphate [ATP] synthase in 5 genes), which carry outoxidative phosphorylation in the mitochondrial inner membrane.
In addition to the novel evidence for the abnormal regulation of nucleargenes coding for mitochondrial proteins, we also confirmed previous evidence14 of decreased expression of the 67-kd isoform of glutamicacid decarboxylase (GAD67), the enzyme synthesizing the inhibitoryneurotransmitter γ-aminobutyric acid (GABA) in bipolar disorder (Table 2). Furthermore, the mRNA codingfor the neuropeptide somatostatin, expressed in the oriens-lacunosum/molecularesubtype of hippocampal interneurons,15 wasmost significantly decreased in all 43 differentially affected genes.
We performed hierarchical clustering using the dChip program to identifysamples with similar expression profiles.16- 17 Tolimit noise and increase the strength of our findings, only genes with amplevariability and present calls were used for clustering (Figure 1A). Variability was set at a standard deviation greaterthan 4% of the mean of the expression value, and genes had to be deemed presentin at least 20% of samples. A total of 216 genes met the criteria. These genesshowed that bipolar disorder samples had similar genetic profiles and clusteredtogether (P = .005).
Hierarchical clustering of samples.A, All genes with a standard deviation higher than 4% of the mean of theirexpression value and present calls in at least 20% of samples were used forclustering (n = 216). Significant clustering of bipolar disorder samples wasobserved (P= .005). B, Genes known to be involved in complexesI through V of the mitochondrial respiratory chain and present in at least20% of samples were used for clustering (n = 72). Significant clustering ofbipolar disorder samples (P= .004) and control samples (P= .02) was observed. Redundant probe sets were excluded from clusteringanalysis. Dark-shaded rectangles indicate bipolar disorder; light-shaded rectangles,schizophrenia; open rectangles, controls; L, lithium carbonate; V, valproicacid; and ?, treatment not known.
To further explore regulation trends in functionally related genes,we used MAPPfinder.12 Of 365 genes that weredown-regulated in bipolar disorder with at least a 1.1-fold difference, P<.02, and more than 50% present calls (false discoveryrate, 7.2%), 326 linked to terms defined by the Gene Ontology Consortium.Of the 6 groups that achieved a z score higher than10, three were associated with mitochondria and 3 with the ATP-dependent processof proteasome degradation (Table 3).MAPPfinder identified 50 mRNA molecules coding for proteins located in themitochondrial inner membrane that were in the Gene Ontology Consortium databaseand HG-U95Av2 array and found 17 (34%) to be decreased in bipolar disorder.Furthermore, the expression of 7 (78%) of 9 genes associated with the proton-transportingATP synthase complex in the inner mitochondrial membrane was decreased inbipolar disorder. The dChip results were verified with Data Mining Tool software,and down-regulation of the same 2 gene families was confirmed with MAPPfinder.
To provide an unbiased review of the regulation of genes involved inenergy metabolism and proteasome degradation, we used GenMAPP to draw mapsof all relevant genes represented in the HG-U95Av2 array.12 Thesemaps revealed that the decreased expression of genes related to mitochondrialfunction was not only pronounced but also widespread (Figure 2 and Figure 3).Similar maps for the schizophrenia group revealed that not a single one ofthe genes listed for oxidative phosphorylation and proteasome degradation,62 and 28 genes, respectively, reached a probability level of P≤.05).
Genes coding for mitochondrialproteins involved in oxidative phosphorylation. The figure includes all genesrepresented on the HG-U95Av2 array (Affymetrix, Santa Clara, Calif) that codefor mitochondrial proteins involved in oxidative phosphorylation. Comparisonof the bipolar disorder group with the control group revealed that most geneswere down-regulated (indicated in blue), some were not significantly changedor had a presence call lower than 60% (indicated in yellow), and none weresignificantly increased. ATP indicates adenosine triphosphate; COX, cytochrome-c oxidase; Cyt, cytochrome; NADH, nicotinamide adenine dinucleotide;OSCP, oligomycin sensitivity–conferring protein; and RISP, Rieske iron-sulfurprotein. A dashed line around a box indicates that the gene was representedon the chip more than once (eg, Cyt C1); if a box has more than 1 color, thevarious representations of the gene on the array met different criteria asindicated by the colors (eg, subunit E). Fold difference is shown to the rightof the box (positive values calculated as bipolar disorder/control; negativevalues calculated as control/bipolar disorder).
Genes coding for proteins involvedin proteasome degradation. The figure includes all genes represented on theHG-U95Av2 array (Affymetrix, Santa Clara, Calif) that code for proteins involvedin proteasome degradation. Comparison of the bipolar disorder group with thecontrol group revealed that most genes were down-regulated (indicated in blue),some were not significantly changed (indicated in yellow), and none were significantlyincreased. See Figure 2 legend for details.
Hierarchical clustering was performed with 72 genes known to be involvedin complexes I through V of the mitochondrial respiratory chain. Significantclustering of bipolar disorder samples (P = .004)and control samples (P = .02) was observed (Figure 1B), demonstrating that these subjectshad a similar profile for genes in the mitochondrial respiratory chain. Nosignificant clustering was observed with valproic acid or lithium carbonatetreatment. In a direct comparison of expression levels of mitochondrial genes,no difference was observed in patients treated with valproic acid or lithiumcompared with those who did not receive the drug.
Because a lower percentage of genes were deemed present in the bipolardisorder group (Table 1), we examinedwhether a lower percentage of gene-present calls was associated with down-regulationof genes in the mitochondrial respiratory pathway, independent of clinicaldiagnosis. We combined control and schizophrenia samples, sorted them accordingto percentage of gene-present calls, compared the 9 samples with the lowerpercentage of gene-present calls (mean ± SD, 42.4% ± 3.1%) withthe 9 samples with a higher percentage of calls (mean ± SD, 47.3% ±0.6%; P<.001), and performed MAPPfinder analysis.Mitochondrial membrane, mitochondrial inner membrane, and proton-transportingATP synthase complex had z scores of 2.7 (ranked236), 0.72 (ranked 608), and 1.6 (ranked 387), respectively. These low z scores (compared with the high z scoresreported in Table 3 for the bipolardisorder group) indicate that the decreased expression of nuclear genes relatedto mitochondrial function cannot be explained solely by the lower percentageof gene-present calls in the bipolar disorder group.
To confirm our finding of decreased expression of nuclear genes involvedin energy metabolism and proteasome degradation in bipolar disorder, we selected4 genes for verification in a real-time quantitative PCR assay: 2 from themitochondrial respiratory chain and 2 proteasome subunits (Figure 4). All 4 mRNA molecules were corrected for an internal controlgene (human filamin A α) and were significantly down-regulated, confirmingthe gene array data.
Real-time quantitative polymerasechain reaction of mitochondrial and protosomal genes in subjects with bipolardisorder and controls. Four genes were selected for verification. A and E,The oligomycin sensitivity–conferring protein (OSCP), a subunit of mitochondrialadenosine triphosphate synthase. B and F, The mitochondrial cytochrome-c oxidase subunit, COX VIIb. C and G, The proteasome α-3 subunit.D and F, The proteasome β-4 subunit. The expression of each gene wasnormalized to human filamin A α. Mean ± SEM is shown. Asteriskindicates significance at P<.05.
To explore whether the decreased expression of these 4 genes was specificfor the hippocampus, we performed the same real-time quantitative PCR analysisin frontal cortex specimens from the same subjects (Figure 4). We found a similar pattern of decreased expression inthe frontal cortex tissue in subjects with bipolar disorder.
Our results provide evidence for the abnormal regulation of nucleargenes coding for mitochondrial proteins in bipolar disorder. In addition,we confirm and extend previous evidence14 ofabnormal gene expression in hippocampal interneurons in bipolar disorder.The decreased expression of GAD67 and somatostatin points to aspecific deficit of the oriens-lacunosum/moleculare subtype of hippocampalinterneurons in bipolar disorder.15 This supportsthe notion that a subset of hippocampal interneurons, located in the stratumoriens and terminating with apical dendrites of principal cells in conjunctionwith perforant pathway afferent fibers, is abnormal in bipolar disorder.
In this article, we focus primarily on the novel evidence for abnormalmitochondrial energy metabolism in bipolar disorder. First, the expressionof genes coding for the enzymatic complexes governing oxidative phosphorylationis decreased in bipolar disorder. Second, the ATP-dependent process of proteasomedegradation is down-regulated at the level of gene expression. This molecularevidence strengthens the hypothesis that decreased pH and high-energy phosphatelevels in bipolar disorder are the result of mitochondrial dysfunction.4
It is unclear how a widespread decrease in the expression of nucleargenes coding for mitochondrial proteins could have been produced in the casesof bipolar disorder reported in this study. Although mutations in both mitochondrial18 and nuclear DNA may contribute to mitochondrial dysfunctionin bipolar disorder, it is unlikely that such mutations could induce the patternof decreased mRNA expression observed in our sample. Possible explanationsof the findings are that the number of mitochondria per neuron is reducedin bipolar disorder or that a subset of neurons with high mitochondrial numbers(eg, GABAergic interneurons) is lost. That mRNA coding for the neuropeptidesomatostatin, expressed in hippocampal interneurons, is also reduced mightsupport this notion of selective neuron loss. Because glial fibrillary acidicprotein mRNA, a marker of gliosis, was not altered in bipolar disorder, neuronaldeath would probably not be a recent event. Alternatively, mechanisms thatcontrol transcription, including the ATP-dependent process of nucleosome remodeling19 or histone acetylation and methylation,20 couldbe involved in widespread changes of gene expression. In this context, itis of interest that lithium and valproic acid, 2 therapeutic agents in thetreatment of bipolar disorder, affect chromatin remodeling. Inositol polyphosphates,targets of lithium can modulate the activities of chromatin-remodeling complexesin vitro.21 The mood-stabilizing drug valproicacid is an inhibitor of histone deacetylase.22 Theinhibition of this enzyme results in a widespread increase in gene expression,including the gene GAD67,23 whichhas been found to be decreased in the hippocampus in bipolar disorder in thisand previous studies.14,24 Itis therefore conceivable that mechanisms of chromatin structuring are affectedin bipolar disorder and are targeted by pharmaceutical compounds effectivein the treatment of this disease.
Despite using all possible measures to avoid the introduction of experimentalbias and although equal amounts of biotinylated RNA were used in all arrays,the percentage of gene-present calls was lower in bipolar disorder. Althoughnot significant, data from the 28S/18S ribosomal RNA ratios and 3′/5′glyceraldehyde-3-phosphate dehydrogenase and β-actin ratios might suggestreduced mRNA quality in the hippocampus in bipolar disorder. Compromised energymetabolism could account for this observation, but we cannot entirely excludethe possibility that factors inherent in postmortem studies and beyond theinvestigators' control might have contributed to reduced RNA quality.
Most subjects with bipolar disorder and all subjects with schizophreniawere treated with neuroleptic medication, which has supportive as well asinhibitory effects on mitochondrial function.25- 27 Ifthe genes for mitochondrial respiration were down-regulated as a result ofantipsychotic drug treatment, this effect should have been more pronouncedin the subjects with schizophrenia, who were treated with higher doses ofantipsychotic drugs. Conversely, if antipsychotic drug treatment up-regulatesgenes for mitochondrial respiration, it could explain why the schizophreniagroup had levels more comparable with controls. Lithium and valproic aciddid not seem to be responsible for the down-regulation of genes because theydid not cluster together (Figure 1B),and lithium had no such effect in an animal study.28 Weconsider this as evidence that the decreased gene expression in our bipolardisorder sample was not due to neuroleptic medication, lithium, or valproicacid.
Recent postmortem studies of schizophrenia have reported that the activityof oxidative enzymes associated with mitochondria, such as the malate-aspartateshuttle system29 and complex IV,30 isalso decreased in the frontal cortex in subjects with schizophrenia. Thissuggests that disturbances in mitochondrial oxidation (at the level of geneexpression, as in our study of subjects with bipolar disorder, or at the levelof enzyme activity, as previously reported in those with schizophrenia) mayplay a broader role in psychotic disorders.
We do not know whether our finding of abnormal gene expression in bipolardisorder is specific to the hippocampus. The results of real-time quantitativePCR analysis of the frontal cortex specimens indicate that the changes reportedin our article are not limited to the hippocampus. Previous studies have demonstratedabnormal gene expression in the hippocampus and cerebral cortex in bipolardisorder,31- 34 butwe are aware of no studies that have systematically examined the expressionof genes coding for mitochondrial proteins in this disease.
It is likely that decreased nuclear gene expression governing oxidativephosphorylation has functional implications. Mitochondrial dysregulation associatedwith decreased oxidative phosporylation shifts metabolism toward anaerobicenergy production via glycolysis, increasing lactate levels and pH and leadingto reactive oxygen species, glutamate excitotoxicity, and apoptosis.35 Similarly, decreased expression of genes coding forproteins of the ubiquitin-proteasome system has functional implications, amongthem an impairment of synapse remodeling.36 Furtherstudies should test the hypothesis that the pronounced and widespread decreaseof mRNA coding for mitochondrial and proteasome proteins leads to abnormalprotein concentration and function. It appears that our finding of a decreasedexpression of genes involved in mitochondrial function and proteasome degradationprovides potential targets for the development of novel drug compounds inthe treatment of bipolar disorder.
Corresponding author and reprints: Christine Konradi, PhD, McLeanHospital, Mailman Research Center, 115 Mill St, Belmont, MA 02478 (e-mail:konradi@mclean .harvard.edu).
Submitted for publication April 30, 2003; accepted October 8, 2003.
This work was supported by grants from the National Institute of MentalHealth (Dr Benes) and the Stanley Foundation (Dr Heckers), Bethesda, Md, andby a gift from Jim and Pat Poitras (Dr Konradi).
We thank Wing Wong, PhD, and Cheng Li, PhD (Department of Biostatistics,Harvard School of Public Health, Boston, Mass), for advice and access to thedChip program, and George Tejada, MS, and the members of the Harvard BrainTissue Resource Center for experimental support.
Country-Specific Mortality and Growth Failure in Infancy and Yound Children and Association With Material Stature
Use interactive graphics and maps to view and sort country-specific infant and early dhildhood mortality and growth failure data and their association with maternal
Thank you for submitting a comment on this article. It will be reviewed by JAMA Psychiatry editors. You will be notified when your comment has been published. Comments should not exceed 500 words of text and 10 references.
Do not submit personal medical questions or information that could identify a specific patient, questions about a particular case, or general inquiries to an author. Only content that has not been published, posted, or submitted elsewhere should be submitted. By submitting this Comment, you and any coauthors transfer copyright to the journal if your Comment is posted.
* = Required Field
Disclosure of Any Conflicts of Interest*
Indicate all relevant conflicts of interest of each author below, including all relevant financial interests, activities, and relationships within the past 3 years including, but not limited to, employment, affiliation, grants or funding, consultancies, honoraria or payment, speakers’ bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued. If all authors have none, check "No potential conflicts or relevant financial interests" in the box below. Please also indicate any funding received in support of this work. The information will be posted with your response.
Register and get free email Table of Contents alerts, saved searches, PowerPoint downloads, CME quizzes, and more
Subscribe for full-text access to content from 1998 forward and a host of useful features
Activate your current subscription (AMA members and current subscribers)
Purchase Online Access to this article for 24 hours
Some tools below are only available to our subscribers or users with an online account.
Download citation file:
Web of Science® Times Cited: 189
Customize your page view by dragging & repositioning the boxes below.
and access these and other features:
Enter your username and email address. We'll send you a link to reset your password.
Enter your username and email address. We'll send instructions on how to reset your password to the email address we have on record.
Athens and Shibboleth are access management services that provide single sign-on to protected resources. They replace the multiple user names and passwords necessary to access subscription-based content with a single user name and password that can be entered once per session. It operates independently of a user's location or IP address. If your institution uses Athens or Shibboleth authentication, please contact your site administrator to receive your user name and password.