a b s t r a c t

Intro: Patients with small intracranial hemorrhage at initial presentation (ICHi) have a relatively uneventful hos- pital course, as compared with larger ICHi. In this study, we tested the null hypothesis that ICHi does not impact the symptom profile of patients with Traumatic brain injury after discharge.

Methods: In this retrospective study, TBI patients over 18 years of age with a head CT at initial presentation and at least one follow-up visit between 2015 and 2018 were included. Those with vascular risk factors, major psychi- atric comorbidities, neurologic disorders, and TBI / CT evidence of ICH within five years were excluded. Patients were stratified based on the presence or absence of ICHi. Symptom profiles were characterized during early (0-3 months post-TBI) and late follow up (4-12 months post-TBI). An adapted 15-question Post-Concussion Symp- tom Scale and a vestibulo-oculomotor (VOM) exam were assessed by a TBI specialist. We compared the age ad- justed clinical symptom profiles between those with and without ICHi.

Results: 69 patients met inclusion/exclusion criteria. 26 (37.8%) had ICHi and 43 (62.32%) did not have ICH. The severity of measured symptoms or VOM findings were not more severe in those with ICHi. Age-adjusted analyses did not show any effect on these outcomes.

Conclusion: ICHi does not impact the symptom profile of patients with TBI in either short or long term.

(C) 2021

1. Intro

traumatic brain injury affects a significant portion of the pop- ulation with over 2.5 million emergency department visits, hospitaliza- tions, and deaths annually in the United States of America [1]. Patients who suffer from TBI may experience a broad range of neurologic deficits that negatively impact a person’s quality of life. Clinical presentation after TBI is non-specific and has been measured using various scales such as the Post-Concussion Symptom Survey (PCSS) to examine phys- ical, cognitive, and psychological functions [2]. TBI of any severity is also known to affect the vestibulo-ocular motor system resulting in symp- toms including nystagmus and impaired gait, which have been reported in over 50% of concussions [3]. A vestibulo-ocular motor (VOM) screen provides objective evidence for TBI as opposed to subjective reported symptoms [4-6]. In addition, when the vestibulo-ocular motor system is not functioning adequately, rapid head movements can elicit symp- toms like dizziness, nausea, and vomiting. These symptoms can occur

* Corresponding author.

E-mail addresses: [email protected] (K. Costenbader), [email protected]

(F. Huda), [email protected] (M. Shand), [email protected] (D. Brown), [email protected] (M. Kraus), [email protected] (R. Taheri).

immediately after TBI and have been reported to indicate a 6-fold in- creased risk of prolonged recovery [3,7]. Dizziness after TBI is the greatest risk factor for a prolonged recovery [8,9].

While some patients with TBI go on to develop lasting neurologic deficits, little is known about predictive factors which correlate with long-term symptoms. Of the many complications attributed to TBI, in- tracranial hemorrhage (ICH) is a potentially devastating consequence which can lead to increased intracranial pressure, with risk of herniation and the secondary effect of toxicity from blood products [16,17]. Several studies have examined the prognostic value of ICH on long-term neuro- logic deficits with varying results. Sadowski-Cron et al., demonstrated an association of ICH with long term neurologic deficits while more re- cent studies by Lee et al., Stenberg et al., and Brown et al. have demon- strated no clear effect [18-21]. As such, there remains a need for further research in the identification of factors that will predict the long term consequences of TBI to manage patient and family expectations and fur- ther improve treatment and rehabilitation in these patients [22]. Under- standing long term outcomes is important to the emergency physician to triage patients appropriately and provide guidance on follow up.

In a previous study, we have shown that patients with a small acute subdural hemorrhage (SDH) have a relatively uneventful hospital course, as compared to those with larger SDH [23]. The objective of

https://doi.org/10.1016/j.ajem.2021.03.053

0735-6757/(C) 2021

image acquisition“>the current study is to determine the impact of SDH as an objective marker of injury on the long-term clinical symptom profile of patients with TBI who follow up in a TBI clinic after discharge from initial hospitalization.

1. Materials and methods
   1. Participants

In this retrospective study, patients with any severity of TBI over 18 years of age with a head CT at the time of initial presentation and with at least one outpatient follow-up visit between 2015 and 2018 were in- cluded. Data was retrieved from the clinical charts of patients who were following with a neuropsychiatrist specializing in TBI at a post-concussion clinic. Those with vascular risk factors, major psychiat- ric comorbidities, neurologic disorders other than TBI, and prior TBI/CT evidence of ICH within five years prior to this study were excluded as these have been shown to change patient outcomes [24,25]. All study protocols were approved by the Institutional Review Board.

• 1. Image acquisition

Axial CT images of the head without contrast were obtained from the hospital department of radiology and reviewed by a board certified neu- roradiologist. Patients were stratified based on the presence or absence of ICHi and the size of SDH hemorrhages was calculated using the ABC/2 method where A is the hemorrhage length, B is the width of hemor- rhage, and C is the number of CT slices with hemorrhage multiplied by slice thickness [26]. To categorize SDH by volume, patients were dichot- omized into three groups. Group A was defined as an SDH volume mea- suring less than 10 cm3 with isolated SDH. Group B was defined as an SDH volume less than 10 cm3 and/or an additional component of ICHi including epidural hematoma, parenchymal contusion, or subarachnoid hemorrhage . Group C was defined as an SDH volume of greater than 10 cm3 with or without an additional component of ICHi. These categories are in keeping with the methodology used in our previous study [23].

• 1. Neurocognitive assessment

We used an adapted post-concussion symptom scale to measure symptoms spanning both vestibular and cognitive domains [2]. The original validated scale by Chen et al. [2] was adapted in a very minor way by a TBI specialist to capture the symptoms that are most useful in the clinical assessment of head injury and concussion patients. The

• 1. Statistical analysis

We first described characteristics of the study population both over- all and by ICHi status using N (%) for categorical variables and mean (SD) for continuous variables. Fisher’s Exact tests were used to compare the difference in patient clinical symptoms by ICHi status for both early and late follow-ups. Additionally, adjusted analyses were performed to control for the effect of age. Furthermore, sensitivity analyses were con- ducted to compare clinical symptoms in patients with category C bleeds and those without ICHi. All analyses were performed using SAS 9.3 (SAS Institute Inc). All tests of significance were 2-sided with ? levels set at 0.05.

1. Results

Overall, 69 patients met study inclusion and exclusion criteria of which 26 (37.8%) had ICHi and 43 (62.32%) did not. The mean age was higher in the ICHi group with P = 0.0086; otherwise, groups were not demographically different Table 1 . At neither early nor late follow up were the severities of measured symptoms or VOM findings more severe in those with ICHi. Further age-adjusted analyses did not show any effect on these outcomes. The volumes of subdural hemor- rhage (SDH) spanned the aforementioned categories with 8 patients (31%) in group A, 11 patients (42%) in group B, and 5 patients (19%) in group C. The average volume of category C bleeds was 22 cm3. In category C there were 3 cases with an additional component of sub- arachnoid hemorrhage and 1 case with epidural hematoma Within Category B there were 9 cases with an additional component of epidural hematoma. Fig. 1 demonstrates an example of a category A ICHi while Fig. 2 demonstrates a category C ICHi in our study. Two of the patients in our study did not have a significant enough SDH component to their bleed and thus volume was not calculated and were not calculated in sensitivity analysis of effect of SDH volume on symptoms. Volume of the SDH did not demonstrate a significant difference in demographic characteristics. Furthermore, when category C bleeds were compared with the patient’s without ICHi, there was no significant difference in number of symptoms experienced at initial or follow up visits. Sensitiv- ity analyses were conducted to study the effect of volume of ICHi on out- comes. Patients in Group A, B, or C did not differ significantly in terms of outcome of symptoms at either initial or follow up visit (see sensitivity and 2 in supplement). A power analysis was not pursued given the ret- rospective nature of our study. Although Power calculations are very

Table 1

Characteristics of TBI study population.

measured symptoms used were those which were available for all the

patients in the study which include headache, nausea, vomiting, balance problems, dizziness, fatigue, sleep changes, irritability, sadness, feeling mentally slow, difficulty concentrating, difficulty remembering, diffi- culty with speech, and visual disturbance. These symptoms are all found on commonly used validated concussion checklists. Each ques- tion domain was evaluated on a 0-6 scale. The scale was then clubbed into 3 categories of mild (0-2), moderate (3-4), and severe (5-6). In ad- dition, a vestibulo-oculomotor screen, including assessment of nystag- mus after rapid head movements, pursuits or saccades on eye tracking, and balance testing, was assessed by a board certified neuro- psychiatrist specializing in TBI in a post-concussion clinic. Presence of each VOM symptom was characterized with a binary scale for presence or absence symptoms. Symptom profiles were characterized during early (0-3 months post-TBI) and/or late follow ups (4-12 months post-TBI). Age adjusted clinical symptom profiles between those pa- tients with ICHi and those without ICHi at each visit were compared. These results were further analyzed to evaluate the impact of the size of SDH.

Characteristic

ICH, N (%)

Yes 26 (37.68)

No 43 (62.32)

Visits, N (%)

0-3 Month Only 27 (39.13)

4-7 Month Only 9 (13.04)

Both 33 (47.83)

Age, Mean (SD) 39.41 (13.63)

Gender, N (%)

Male 34 (49.28)

Female 35 (50.72)

Depression

Yes 13 (18.84)

No 56 (18.84)

Other CT Findings

SAH 13 (18.84)

PHI 3 (4.35)

EDH 2 (2.90)

PHI and EDH 1 (1.45)

None 50 (72.46)

Fig. 1. Category A ICHi.

Fig. 2. Category C ICHi.

useful for study design proposes, their usefulness for retrospective stud- ies is less accepted [42,43].

1. Discussion

In this study we tested the hypothesis that long-term morbidity after TBI, including both clinical symptoms and VOM signs, could be stratified by the presence of ICH and size of SDH. Demographic analysis demon- strated that older age was associated with a higher rate of ICH. This find- ing makes sense given that older populations have more brain atrophy, which places more tension on the bridging veins, and are therefore more susceptible to ICH due to tearing of bridging veins [27].

The variation of hemorrhages in our population appears similarly distributed to other studies. In one larger study with about 14,000 pa- tients with TBI, 53% of patient’s had no hemorrhage. A small subset of patient’s, 4200 had SDH of which 26% were considered large and 28% were considered small (criteria for size estimate was not stated in this study) [44].

In our study, the presence of ICHi alone did not demonstrate any sig- nificant prognostic value on long-term outcomes in patients with TBI. After stratifying by size of SDH, there was no difference in the symptom- atic profile between those with or without ICH suggesting that neither presence nor size of SDH were predictive of prolonged experience of clinical symptoms at one year in our study population.

Similar studies have reported conflicting results with several studies demonstrating that certain CT findings can predict long-term outcomes [18,28], while more recent studies have failed to find any predictive value of CT or MRI for neurocognitive outcomes at one year [19]. One large multicenter study used the Marshall head CT classification system and assessed for a relative change in neurocognitive outcomes based on severity of injury. The results found no predictive effect of CT on long- term neurocognitive outcomes after injury at any point in the study [20]. It is notable to mention that the aforementioned study used an adapted Marshal head CT scale that purposefully excluded the volume of lesions as the measurement was not consistently available in their ra- diology reports. In another study specifically looking at severe TBI, there was again no association found between head CT characteristics and clinical outcomes as measured by the Glasgow Outcome Scale Extended (GOSE) at either three months or one year. Again, this study did not re- port or measure the volume of ICH in its use of the Marshal head CT [21]. Variation in these studies may be due to the differences in outcomes measured and/or cohort variability. Another possibility for this variation is that the deficits experienced after TBI are microstructural in nature or represent individual variation in brain connections, i.e. connectomes which may be better evaluated through more advanced imaging tech- niques. There has been some investigation into the prognostic value of diffusion tensor imaging (DTI) in the semi Acute stage (9-14 days) after mild TBI which showed that DTI was superior to CT in predicting three -and six-month outcomes [30]. Another study suggests that neurite orientation dispersion and density imaging (NODDI) may be more sensitive than other DTI metrics for prognosis [31]. Future work in diffusion MRI and fMRI may better characterize an association be-

tween radiologic findings and clinical outcomes.

Previously, other studies have described VOM symptoms, specifi- cally dizziness, as being associated with a protracted recovery in pa- tients with mild TBI without ICHi [8,32-36]. In healthy people without brain injury, the vestibulo-oculomotor reflex coordinates eye move- ments and helps maintain balance during movements. This VOM func- tion is important for every day activity and more complicated tasks like sports. Dysfunction typically manifests as dizziness or disequilib- rium. These symptoms generally have been identified as a risk factor for prolonged recovery from TBI. Therefore, this study aimed to identify whether volume of SDH was associated with prolonged or worse symp- toms compared to those with small SDH or no ICHi at all [37]. The results of this study found no association with the presence of ICHi and fre- quency of VOM symptoms, suggesting that ICHi may not be appropriate

in determining the prognosis of VOM symptoms in patients with TBI. While VOM symptoms have been associated with prolonged recovery of TBI symptoms in previous studies, these symptoms do not seem to be associated with the presence or size of ICH. This may also suggest that the etiology of VOM symptoms might be attributed to inner ear, vi- sual, or Spinal injury which are part of the VOM system, instead of brain injury. It is also possible that another imaging modality is required to better predict these outcomes.

The findings in our current study have clinical relevance as patients and their families rely on their medical team to provide both diagnosis and prognosis of their injury. It is of importance to the emergency phy- sician as most patients with TBI do not follow up with a TBI clinic and may be more likely to return to the emergency department for other reasons. This study contributes to the previous literature which find limited utility of head CT findings to predict long term symptomatic out- comes [19-21]. Ultimately, this highlights the need for more sophisti- cated studies which examine functional data to provide a prognosis for those who suffer from TBI [30,38]. Better prognostication will allow better triage of patients and target rehabilitation for those with worse prognosis.

1. Limitations

There are several limitations to our study. We were only able to ob- tain a small study sample given that a significant portion of patients who initially presented to our emergency department for TBI were lost to follow up. Although our findings should be taken in context of our limited sample size, other studies with comparable sample sizes proved negative with respect to prognostic value of CT on patient out- comes [18-19,21].

While CT scans are generally the modality of choice in TBI, these are a static image which may miss progression of hemorrhage and thus can underestimate the true extent of pathologies after TBI. This may poten- tially limit our ability to detect Significant injuries contributing to patient’s who didn’t experience ICHi but may have had other injuries such as Diffuse axonal injury [40] Additionally, the clinical symptoms experienced after TBI are not only influenced by trauma related factors, but also a variety of other factors [41]. For example, TBI recovery is influ- enced by motivation, psychosocial support, clinical follow up, and reha- bilitation which can have a positive influence on recovery and may have interfered with our results.

Finally, the results reported only relate to a population that survived TBI and were seen in a post-acute care clinic. No extrapolation can be made to patients seen with ICHi in the acute setting or those without close follow up or rehabilitation.

This study can be strengthened in the future by including a larger sample population, the use of more advanced imaging like MRI or DTI, and additional parameters for CT findings, such as location of hemor- rhage or automatic quantification of hemorrhage with the assistance of artificial intelligence algorithms. Future studies should also incorpo- rate emergency room physicians to perform symptomatic screening on patients with TBI and advanced imaging as these patients may be more likely to return to the emergency department for other reasons in- stead of presenting to a TBI clinic.

1. Conclusion

In summary, we assessed two new approaches to evaluate the prog- nostic value of ICH, specifically SDH, at the initial presentation of TBI. First, we assessed the use of SDH volume stratification based on initial head CTs performed in our emergency department for prognostication of long-term patient reported clinical symptoms as measured by the PCSS. Second, we sought to establish the relationship between ICH and VOM signs in patients with TBI. In conclusion, neither the presence of ICH nor size of SDH seem to impact the clinical symptom profile of pa- tients with TBI in either the short or long term. At no point in our study,

was the PCSS or VOM system function worse in patients with larger size of SDH. Efforts to prognosticate patients with TBI should switch from CT evaluation to more advanced studies to evaluate function including DTI. Ultimately, better data on which to base prognosis after TBI will help guide patient and family expectations, and improve the emergency physician’s ability to triage patients on discharge from the emergency department.

Conflicts of interest

None.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi. org/10.1016/j.ajem.2021.03.053.

References

1. Centers for Disease Control and Prevention. Injury prevention & control: traumatic brain injury & concussion. http://www.cdc.gov/traumaticbraininjury/data/rates. html.
2. Chen J, Johnston KM. Collie a, et alA validation of the post concussion symptom scale in the assessment of complex concussion using cognitive testing and functional MRI Journal of Neurology. Neurosurg Psychiatry. 2007;78:1231-8.
3. Lau BC, Kontos AP, Collins MW, Mucha A, Lovell MR. Which on-field signs/symptoms predict pro- tracted recovery from sport related concussion among high school foot- ball players. Am J Sports Med. 2011;39:2311-8.
4. Marcus HJ, Paine M, Wolstenholme S, Collins K, Marroney N, Arshad Q, et al. Vestib- ular dysfunction in acute traumatic brain injury. Journal of Neurology. 2019;266 (10):2430-3.
5. Mucha A, Collins MW, Elbin RJ, et al. A brief vestibular/ocular motor screening (VOMS) assessment to evaluate concussions: preliminary findings. Am J Sports Med. 2014;42(10):2479-86.
6. Wallace B, Lifshitz J. Traumatic brain injury and vestibulo-ocular function: current challenges and future prospect. Eye Brain. 2016;8:153-64.
7. McCrory P, Meeuwisse W, Dvorak J, et al. Consensus statement on concussion in sport-the 5th international conference on concussion in sport held in Berlin. Octo- ber 2016 Br J Sports Med. 2017;51:838-47.
8. Suh M, Kolster R, Sarkar R, Ghajar J. Deficits in predictive smooth pursuit after mild traumatic brain injury. Neurosci Lett. 2006;401:108-13.
9. Davies RA, Luxon LM. Dizziness following head injury: a neurotological study. J Neurol. 1995;242:222-30.
10. Lok J, Leung W, Murphy S, Butler W, Noviski N, Lo EH. Intracranial hemorrhage: mechanisms of secondary brain injury. Acta Neurochir Suppl. 2011;111:63-9. https://doi.org/10.1007/978-3-7091-0693-8_11.
11. Mckee AC, Daneshvar DH. The neuropathology of traumatic brain injury. Handb Clin Neurol. 2015;127:45-66. https://doi.org/10.1016/B978-0-444-52892-6.00004-0.
12. Sadowski-Cron Charlotte, Schneider Jorg, Senn Pascal, Radanov Bogdan P, Ballinari Pietro, Zimmermann Heinz. Patients with mild traumatic brain injury: immediate and long-term outcome compared to intra-cranial injuries on CT scan. Brain Inj. 2006;20(11):1131-7. https://doi.org/10.1080/02699050600832569.
13. Lee Hana, Wintermark Max, Gean Alisa D, Ghajar Jamshid, Manley Geoffrey T, Mukherjee Pratik. Focal lesions in acute mild traumatic brain injury and neurocognitive outcome: CT versus 3T MRI. J Neurotrauma. Sep 2008;25(9): 1049-56.
14. Brown Allen W, Pretz Christopher R, Bell Kathleen R, Hammond Flora M, Arciniegas David B, Bodien Yelena G, et al. Predictive utility of an adapted Marshall head CT classification scheme after traumatic brain injury. Brain Inj. 2019. https://doi.org/ 10.1080/02699052.2019.1566970.
15. Stenberg Maud, Koskinen Lars-Owe D, Jonasson Per, Levi Richard, Stalnacke Britt- Marie. Computed tomography and clinical outcome in patients with severe trau- matic brain injury. Brain Inj. 2017;31(3):351-8. https://doi.org/10.1080/02699052. 2016.1261303.
16. Saatman KE, Duhaime AC, Bullock R, Maas AI, Valadka A, Manley GT. Classification of traumatic brain injury for targeted therapies, Workshop Scientific Team and Advi- sory Panel Members. J Neurotrauma. 2008 Jul;25(7):719-38.
17. Albertine P, Borofsky S, Brown D, Patel S, Lee W, Caputy A, et al. Small subdural hem- orrhages: is routine intensive care unit admission necessary? Am J Emerg Med. 2016;34(3):521-4. https://doi.org/10.1016/j.ajem.2015.12.035.
18. Dams-O’Connor K, Spielman L, Singh A, et al. The impact of previous traumatic brain injury on health and functioning: a TRACK-TBI study. J Neurotrauma. 2013;30(24): 2014-20. https://doi.org/10.1089/neu.2013.3049.
19. Yue JK, Cnossen MC, Winkler EA, et al. Pre-injury comorbidities are associated with functional impairment and post-concussive symptoms at 3- and 6-months after mild traumatic brain injury: a TRACK-TBI study. Front Neurol. 2019;10:343 Pub- lished 2019 Apr 9 https://doi.org/10.3389/fneur.2019.00343.
20. Gebel JM, Sila CA, Sloan MA, Granger CB, Weisenberger JP, Green CL, et al. Compar- ison of the ABC/2 estimation technique to computer-assisted volumetric analysis of

intraparenchymal and subdural hematomas complicating the GUSTO-1 trial. Stroke. 1998;29(9):1799-801.

1. Armao DM, Bouldin TW. Pathology of the nervous system. In: Reisner HM, editor. Pathology: a modern case study, 2e. New York, NY: McGraw-Hill; 2020http:// accessmedicine.mhmedical.com.proxygw.wrlc.org/content.aspx?bookid=274 8§ionid=230845479.
2. Maas AI, Steyerberg EW, Butcher I, Dammers R, Lu J, Marmarou A, et al. Prognostic value of computerized tomography scan characteristics in traumatic brain injury: re- sults from the IMPACT study. J Neurotrauma. 2007;24(2):303-14.
3. Yuh Esther L, Cooper S, Mukherjee P, Yue John K, Lingsma H, Gordon W, et al. Diffu- sion tensor imaging for outcome prediction in mild traumatic brain injury: a TRACK- TBI Study. (report). J Neurotrauma. 2014;31(17):1457-77. https://doi.org/10.1089/ neu.2013.3171.
4. Palacios EM, Owen JP, Yuh EL, Wang MB, Vassar MJ, Ferguson AR, et al. The evolution of white matter changes after mild traumatic brain injury: A DTI and NODDI study. bioRxiv. 2018. https://doi.org/10.1101/345629.
5. Wright WG, Tierney RT, McDevitt J. Visual-vestibular processing deficits in mild traumatic brain injury. J Vestibular Res. 2017;27:27-37. https://doi.org/10.3233/ VES-170607.
6. Lau BC, Kontos AP, Collins MW, Mucha A, Lovell MR. Which on-field signs/symptoms predict protracted recovery from sport-related concussion among high school foot- ball players? Am J Sports Med. 2011;39(11):2311-8.
7. Denby E, Murphy D, Busuttil W, Sakel M, Wilkinson D, et al. Neuropsychiatric Out- comes in UK Military Veterans With Mild Traumatic Brain Injury and Vestibular Dys- function. J Head Trauma Rehabil. 2020;35(1):57-65.
8. Gottshall K, Drake A, Gray N, McDonald E, Hoffer ME. Objective vestibular tests as outcome measures in head injury patients. Laryngoscope. 2003;113:1746-50.
9. Davies RA, Luxon LM. Dizziness following head a neurotological study. J Neurol. 1995;242:222-30.
10. Wallace B, Lifshitz J. Traumatic brain injury and vestibulo-ocular function: current challenges and future prospects. Eye Brain. 2016;8:153-64 Retrieved from https:// doaj.org/article/e8d2c43e894a44858269bf6c7ca56d4b.
11. Yuh EL, Mukherjee P, Lingsma HF, Yue JK, Ferguson AR, Gordon WA, et al. Magnetic resonance imaging improves 3-month outcome prediction in mild traumatic brain injury. Ann Neurol. 2013 Feb;73(2):224-35.
12. Mittl R, Grossman I, Hiehle J, Hurst R, et al. Prevalence of MR evidence of diffuse ax- onal injury in patients with mild head injury and normal head CT findings. Am J Neuroradiol. 1994;15:1583-9.
13. Ponsford J, Willmott C, Rothwell A, Cameron P, Kelly AM, Nelms R, et al. Factors influencing outcome following mild traumatic brain injury in adults. J Int Neuropsychol Soc. 2000;6:568-79.
14. Thomas L. Retrospective power analysis. Conserv Biol. 1997;11(1):276-80.
15. Hoenig and Heisey. The abuse of power. Am Stat. 2001;55(1):19-24.
16. Perel P, Roberts I, Bouamra O, Woodford M, Mooney J, Lecky F. Intracranial bleeding in patients with traumatic brain injury: a prognostic study. BMC Emerg Med. 2009 Aug 3;9:15. https://doi.org/10.1186/1471-227X-9-15 PMID: 19650902; PMCID: PMC2735732.