Address for Correspondence Clive H. Wilder-Smith, Brain-Gut Research Group, Bubenbergplatz 11, CH-3011 Bern, Switzerland. Tel: +41 31 312 3737; fax: +41 31 312 3770; e-mail: firstname.lastname@example.org
Background Visceral hypersensitivity is one of the proposed underlying mechanisms in functional dyspepsia (FD). It is not clear whether visceral hypersensitivity in FD is a manifestation of a central sensitization also encompassing somatic sensitization. Transient receptor potential vanilloid-1 (TRPV1) pathways are involved in gastric mechanosensory physiology and the TRPV1 receptor agonist, capsaicin, has been used as a chemical stimulant.
Methods In this double-blind, randomized study we evaluated both visceral and somatic sensory function in 34 FD patients and 42 healthy controls using quantitative sensory testing. Visceral pain sensitivity was assessed using a validated gastric pain model with oral capsaicin capsule titration and somatic pain sensitivity was determined by foot heat and hand electric stimulation.
Key Results The median capsaicin dose required to attain moderate pain was 0.5mg in FD and 1mg in controls (P = 0.03). At these doses, mean pain intensities on a 0–100 visual analog scale were greater for FD than controls [56.9 (95% confidence intervals, 52.2–61.5) vs 45.1 (41.6–48.6), resp.] (P = 0.005). Overall, mean somatic sensory and pain thresholds were similar in FD and control groups, but in a subgroup of FD pain hypersensitivity was seen on the hand and on the foot at different stimulation thresholds.
Conclusions & Inferences A majority of patients with FD have visceral chemo-hypersensitivity involving TRPV1 pathways. A substantial subgroup also has somatic hypersensitivity as evidence of central sensitization.
Visceral hypersensitivity is a proposed underlying mechanism in Functional Gastrointestinal Disorders (FGIDs), for which there is considerable evidence especially in Irritable Bowel Syndrome (IBS).1 Many patients with functional dyspepsia (FD) have evidence of visceral sensory abnormalities, including lower discomfort or pain thresholds to gastric distension.2–7 The mechanisms behind these sensory changes remain unclear, but may include impaired gastric accommodation, abnormal fundic phasic contractions, sensitization following inflammation or infection, stress or psychosocial comorbidity, and modifications in central sensory processing.8–13
Visceral sensory testing in FD has commonly been performed using distension with a barostat-controlled intragastric bag. However, the tolerability is poor and there is contamination with a range of intense and non-physiologic stimuli.14 Besides mechanical distension, chemical or nutrient-induced stimuli could precipitate the postprandial symptoms defining FD.15–19 Chemical stimulation may be a more physiologic and multimodal yet less invasive means of assessing gastric sensory changes than distension. Upregulation and sensitization of the transient receptor potential vanilloid-1 (TRPV1) pathway likely plays an important role in chemical hyperalgesia in the upper as well as the lower GI tract.20 Capsaicin, the pungent natural ingredient of red chili pepper, is a TRPV1 agonist, exciting afferent C-fiber neurons and causing burning discomfort or pain.21 TRPV1 is not only involved in chemosensing, but also in mechanosensory transduction in the GI tract.20,22,23 Therefore, in contrast with barostat distension, TRPV1-stimulation by capsaicin is likely to represent a polymodal stimulation involving several physiologic pathways central to sensitization.
Various forms of capsaicin challenge have been used to study visceral sensation in controls and in FGID, including infusion of red pepper or capsaicin solution by oral and intestinal intubation, ingestion of a spicy meal or of a fixed-dose of capsaicin.24–29 However, standardized quantitative sensory testing with a dose-response paradigm for demonstration of sensitization in FGID has not been performed with capsaicin, although increased non-painful perception following a fixed-dose of capsaicin has been shown in FD by Hammer et al.20,26,30 Titration to adequate pain intensity for activation of the high-threshold C-fibers involved in TRPV1 pathways is important in the study of sensitization.31,32 We therefore recently developed a gastric pain model using individual oral titration of capsaicin in healthy volunteers.33 The range of pain and non-pain sensory characteristics elicited during capsaicin stimulation corresponds well with those seen clinically in FD. However, the discriminative nature of this new assay has not been tested in FD.
Generalized or central sensitization, i.e. somatic as well as visceral hypersensitivity, has been demonstrated in IBS and non-cardiac chest pain.1,34–38 To the best of our knowledge central sensitization has not been assessed in FD, but somatization (increased generalized somatic symptom reporting), which may also be mediated by central sensitization, is an important risk factor for impairment in health-related quality of life in both FD and IBS.39,40 It is currently not clear whether gastric hypersensitivity in FD is a manifestation of a more generalized central hypersensitivity. The main aims of this study were to compare visceral and somatic pain sensitivity in FD and healthy controls using a quantitative sensory testing paradigm. Our hypothesis was that the majority of patients with FD would show both visceral and somatic hypersensitivity compared to controls. The coexistence of visceral and somatic sensitization is taken as evidence of central generalized sensitization, which may represent an underlying disease mechanism in FD.
Materials and Methods
This was a prospective, randomized, double-blind, nocebo-controlled study performed in FD patients and healthy controls at the National University Hospital (NUH) of Singapore, from September 2008 until September 2011.
Thirty-four male and female FD patients diagnosed according to the Rome III criteria,41,42 aged between 21 and 70 years were recruited from the NUH Department of Gastroenterology or by public advertising. For inclusion, abdominal discomfort or pain had to be the patient’s most prominent symptom and all dyspepsia-related, analgesic, and psychotropic medication had to have been discontinued for at least 2 weeks before the beginning of the study. Forty-two male and female healthy age-matched volunteers were recruited by word-of-mouth. The main exclusion criteria were significant gastrointestinal (GI) symptoms or underlying chronic GI conditions. In both groups a history of other significant systemic illness, including cardiovascular, psychiatric, neurological, and endocrine diseases, any other chronic or acute pain syndromes, bowel resections or abdominal operations except appendectomy, ongoing treatment with any drugs, or complementary medication within the last 14 days before the start of the study, pregnancy or lactation, and absence of written informed consent led to exclusion from the study.
Ethics approval and consideration
The study protocol was approved by the Institutional Review Board and Domain Specific Review Board of the National Healthcare Group, Singapore. All participants gave written informed consent and were remunerated for their participation. The study was registered at Clinical Trials.gov as ID NCT00693407.
Subjects were seen on two occasions. At the first visit, written informed consent was obtained, the Leeds Dyspepsia Questionnaire,43 Hospital Anxiety and Depression (HAD) scale,44 Perceived Stress scale (PSS)-10,45 and a standard dietary questionnaire46 were completed and subjects were familiarized with the study procedures. Patients then completed a 2-week observation run-in period for prospective assessment of symptoms and pain. During this period, an anchored visual analog scale (VAS: 0 = none, 100 = maximum bearable) was used by the patients to record their average abdominal pain and discomfort once a week. For study inclusion, all the inclusion criteria had to be fulfilled and the average pain or discomfort intensity during the run-in period had to be greater than 30. Subsequently, the below experimental procedures were performed after an overnight fast. Gastric and somatic sensory tests were performed at the same time of day in all subjects and within 7–14 days after the end of the menstrual period in women to minimize diurnal and cyclical influences. Identical instructions were given before each test.
Visceral sensory testing
This procedure has been described previously and validated by our group.33 After fasting overnight, subjects were randomized by computer-generated list in one block to oral dosing with identical capsules containing either capsaicin 0.5 mg (Sigma-Aldrich, Singapore) or lactose monohydrate 0.5 mg (NUH Pharmacy, Singapore) as nocebo with 100 ml of water in a double-blind, cross-over fashion. A nocebo is defined as an inert substance given blindly for induction of pain. Following ingestion, abdominal pain was rated every minute on a 100 mm horizontal VAS of 0–100, where 0 = no pain and 100 = worst pain bearable. If pain scores did not reach a minimum level of moderate pain (VAS 30–54,47) for at least five successive minutes within 15 min after capsule ingestion, a further capsule of the same content was swallowed. This was repeated until moderate pain of at least 5 min duration was reported or a maximum number of eight capsules were ingested. Any adverse events were recorded during this titration process. After conclusion of the titration and rating, test subjects drank 500 ml of still water to wash out any remaining capsaicin or nocebo. Two hours later, the same dose-titration procedure was completed with the alternative capsule, this wash-out interval having previously been shown to be adequate for exclusion of any residual effect.21,33 The quality and intensity of gastric sensations other than pain were also recorded after each session. If at any time during the study the pain intensity was above 80 on the VAS, pain was immediately eliminated by drinking of 500 ml of water. Sensory data from these patients were included for analysis up to this point.
Quantitative somatic sensory testing
Subjects were seated in a comfortable upright position for the foot heat and hand electrical sensory testing (see below). To exclude order effects, subjects with an odd subject number received the foot test first and subjects with an even subject number began with the hand stimulation. There was a 10-min break between the two tests.
Foot heat pain stimulation The contact thermode of the Medoc Pathway ATS/CHEPS system (Ramat Yishai, Israel) with a contact area of 9 cm2 was secured on the back of the right foot with an elastic strap and first heat sensation, first pain, and pain tolerance thresholds were determined using an ascending method of limits (AML) protocol with incremental steps of 1 °C and 30 s duration, beginning at 36 °C and with a cut-off maximum temperature of 49 °C. Pain ratings were recorded by subjects on a visual analog scale (VAS graded 0–100) immediately after each stimulation. The stimulation temperature giving rise to a moderate pain intensity of VAS 30–54 was determined from the AML. After a 5-min break, tonic heat stimulation with the determined moderate pain intensity was applied for 30 s and subjects rated the pain intensity by VAS. Subjects were blinded to the temperature used, but were informed the thermode temperature could rise, remain stable, or decrease.
Hand electrical stimulation After degreasing of the skin with ether, ECG gel electrodes connected to the Pajunk MultiStim Vario nerve stimulator (Times Medic Pte Ltd, Singapore) were applied on the thenar eminence of the right hand 2 cm apart. First sensation, first pain, and pain tolerance electrical thresholds were determined using an AML protocol with incremental steps of 0.5 mA (constant current 1 ms stimulation at 100 Hz) of 30 s duration. The maximum cut-off current was 60 mA. Pain ratings were recorded by VAS immediately after each stimulation. The current inducing moderate pain was determined from the AML. After a 5-min break, tonic electrical stimulation with the determined moderate pain intensity was applied for 30 s and subjects rated pain intensity by VAS. Subjects were blinded to the current magnitude used and were informed the current could increase, remain stable, or decrease.
Data and statistical analysis
Statistical analyses were performed using the SPSS software (Version 17.0 for Windows, Chicago, IL, USA). The primary study variable was the number of capsules required by FD patients and controls to induce moderate pain lasting at least 5 min, with moderate pain referring to VAS scores between 30 and 54.47 Further outcome variables were the comparison between FD and control groups of the average gastric pain intensity within 5 min after attaining the target of moderate pain, and the somatic sensation and pain thresholds. Functional dyspepsia patients were classified as viscerally or somatically hypersensitive based on their titrated thresholds below the lower 95% confidence interval (CI) of healthy controls as the threshold limits for hypersensitivity, as suggested in the literature.1,48,49 The average amount of chili consumed per month was quantified using the questionnaire by Kang et al.46 Continuous variables are expressed as arithmetic means with 95% confidence intervals, in the case of a normal distribution. Median values with interquartile ranges are given for data with a non-normal distribution and categorical variables are expressed as frequencies and percentages. The independent-samples t-test or Fisher’s Exact test were used for the group comparisons between patients and controls, as appropriate. The actual pain intensity was compared in patients requiring one or more capsules to achieve the titration end point of moderate pain using a one-way anova. Correlations between the pain stimulation intensity and the ratings of stress, anxiety, or depression were evaluated by Pearson’s and Spearman’s rank correlation coefficient. A two-tailed P value <0.05 was considered statistically significant.
Characterization of participants
The demographic data with further characteristics of all the participants are listed in Table 1. Functional dyspepsia patients showed similar age, gender, race, and lifestyle factors as controls. Only a small minority in both groups avoided chili meals, due to a personal dislike for chili. As expected, patients with FD demonstrated higher frequency of various upper GI symptoms and pain scores (all P < 0.001). Baseline HAD anxiety subset scores were also higher in the FD patients compared with the healthy subjects (P < 0.001). Both groups showed similar stress scores, which were higher than the established Caucasian community norm of 12, but comparable to the average score of 18 in a healthy Asian cohort.45,50
Table 1. Characteristics of participants
*: Definitions in the Leed’s Dyspepsia Questionnaire: Indigestion: a pain or discomfort in the upper abdomen; heartburn: a burning feeling behind the breastbone; regurgitation: an acid taste coming up into the mouth from the stomach; nausea: a feeling of sickness without actually being sick. 1 = Not at all, 2 = Less than once a month, 3 = Between once a month and once a week, 4 = Between once a week and once a day, 5 = Once a day or more. 1Stress level scores range from 0 to 40. 2Scores of anxiety and depression range from 0 to 21, with <7 defined as the normal range. One chili unit equals to the equivalent of one teaspoon of chili. IQR, interquartile range.
Number of subjects
Age (95% CI): years
Female subjects: number (%)
Chinese subjects: number (%)
*Frequency of the symptom over last 2 months, Leeds Dyspepsia Questionnaire (Median IQR)
FD pain scores over last 2 weeks: mean (95% CI)
FD discomfort scores over last 2 weeks: mean (95% CI)
1Stress scores, PSS Questionnaire: mean (95% CI)
2Anxiety scores, HAD Questionnaire: mean (95% CI)
3Depression scores, HAD questionnaire: mean (95% CI)
Regular chili ingestion: number (%)
Chili units eaten per month: mean (95% CI)
Smokers: number (%)
Regular alcohol consumers: number (%)
Regular caffeine consumers: number (%)
Visceral sensory function in FD and controls
Moderate pain was successfully achieved with the standardized capsaicin titration in 33 of the 34 (97%) patients with FD and in 38 of the 42 (91%) healthy participants (P = 0.37). The endpoint of moderate pain was reached with a median of one capsule of capsaicin (IQR 1-1) in FD patients and with two capsules in controls (P = 0.03). Seventy-six percent (25/33) of FD patients and 50% (19/38) of control subjects attained moderate pain with one capsaicin capsule (P = 0.04; Fig. 1). In the FD group, the one male (3%) who failed to reach the target of moderate pain attained a maximum VAS score of 20 despite the full dose of eight capsaicin capsules. In the control group, of the four subjects (10%) who failed to reach moderate pain intensity of 5 min duration, two were males, reaching only mild pain with VAS scores of 4 and 23, respectively, despite consuming eight capsaicin capsules, and two were females reporting strong nausea before pain was attained, who consequently dropped out of the study.
The induced gastric pain was more intense in FD than in controls. At the titrated pain threshold, the median pain intensity was 24.9 (18.7–31.1) in FD and 16.1 (12.1–20.1) in controls (P = 0.013). The median VAS intensity of pain in the 5 min after attaining the threshold of moderate pain was 56.9 (52.2–61.5) in FD and 45.1 (41.6–48.6) in controls (P = 0.005). After the onset of moderate pain, pain persisted for a similar length of time in both FD [8.0 (6.3–9.6) min] and controls [8.3 (6.6–9.9) min; P = 0.80]. Among subjects who achieved moderate pain with one capsule only, FD showed more intense pain than controls at the first pain threshold (VAS mean 29.5 in FD vs 18.5 in controls, P = 0.014) and also in the 5 min following the onset of moderate pain (VAS mean 60.0 in FD vs 49.6 in controls, P = 0.044, Fig. 2). The capsaicin dose–pain response relationship in FD patients and controls are shown in Fig. 2. Twenty-five of 34 (74%) FD patients were viscerally hypersensitive compared to controls, i.e. they required less than the controls’ lower 95% CI of the capsaicin capsule number (i.e. 1.5 capsules) to achieve moderate pain.
Patients requiring one capsaicin capsule to achieve moderate pain did not differ in demographics and clinical characteristics from those requiring two or more capsules, except for a higher prevalence of regurgitation (P = 0.02). One of 42 controls (2%) compared with 4 of 34 FD patients (12%; P = 0.17) experienced moderate pain with nocebo titration (nocebo responders). The control nocebo responder was a male who attained a VAS intensity of 33 after ingesting two capsules. The FD nocebo responders were four females, who reported moderate to severe pain with an average maximal VAS intensity of 58.5 (43.6–73.4). Two had ingested one nocebo capsule and one each two and three nocebo capsules. Among all nocebo responders, the median time (IQR) to moderate pain was 20 (8–23) min compared to the latency of 8 (7–12) min during the capsaicin titration (P = 0.068).
Functional dyspepsia and controls showed a similar incidence of non-painful gastric sensations during capsaicin titration. The most prominent symptoms were gastric burning and fullness, followed by abdominal cramps, stinging, and nausea.
Somatic sensory function in FD and controls
Overall, somatic sensory function was similar in FD patients and controls, with no significant group differences in titrated stimulation intensities at sensation and pain thresholds, nor at moderate pain (Table 2). The first sensation thresholds in FD and controls subjects were 1.1 mA (0.9–1.3) and 1.0 mA (0.9–1.2) on the hand (resp.), and 39.0 °C (38.2–40.0) and 38.2 °C (37.6–38.7) on the foot (resp.), with no significant differences between the groups. The pain intensities during the titrated moderate pain stimulation on the hand and foot were similar between FD patients and controls except for a greater pain intensity at tolerance threshold in hand stimulations in FD patients compared to the controls (Table 2). The hand stimulation intensity at pain tolerance threshold was negatively correlated with the rating of depression (r = −0.41, P = 0.044) in FD patients, there was no other significant correlations between the stimulation intensity or pain VAS and the ratings of stress, anxiety or depression. A subgroup of patients displayed somatic hypersensitivity with stimulation intensities for first sensation and for moderate pain below the lower 95% CI of controls (Table 2 and Fig. 3). A majority of patients (≥75%) with somatic hypersensitivity also showed visceral pain hypersensitivity (Fig. 3). Conversely, a significant number of patients with visceral hypersensitivity (6–16/25, 24–64%) also displayed somatic hypersensitivity (Fig. 3).
Table 2. Somatic (hand electrical and foot heat) stimulation intensities and pain VAS at titrated thresholds in patients with FD and healthy controls. 1st pain: threshold when subjects first felt pain; moderate pain: threshold when subjects felt a pain intensity of VAS 30–54; pain tolerance: threshold when subjects requested termination of stimulation due to upper pain limit. Means and 95% CI are shown
Titration of somatic stimulation thresholds
Pain intensity of somatic stimulation(VAS, 0–100)
Controls (n = 42)
FD (n = 34)
Controls (n = 42)
FD (n = 34)
Hand electrical stimulation (mA)
Foot heat (°C)
In this study, visceral hypersensitivity was demonstrated in FD patients as a group using a simple and standardized individual capsaicin titration paradigm. This was reflected in the lower doses required to induce moderate pain, as well as higher dose-pain responses. Seventy-four percent of patients had visceral hypersensitivity, falling outside the 95% CI of healthy control values. Overall, patients with FD were not more sensitive to somatic stimulation than controls. However, somatic hypersensitivity was evident in approximately 60% of patients on the upper and 30% on the lower extremity. Most patients with somatic hypersensitivity had coexisting visceral hypersensitivity.
Central sensitization as an underlying disease mechanism has not been extensively studied in FD. It can be defined as a shift in the balance between endogenous pain inhibitory and facilitatory mechanisms, which may be driven both by central nervous system processes as well as peripheral input.38 In IBS, central sensitization exists in a majority of patients, demonstrated by both visceral and somatic hypersensitivity.1,34–36 Indirect evidence of nervous system plasticity associated with central sensitization in IBS and FD has been presented in several brain imaging studies.13,51–55
Visceral hypersensitivity was clearly shown in a majority of patients with FD in this study. Using a previously validated double-blind individual capsaicin titration paradigm, patients with FD required a significantly lower capsaicin dose to reach moderate gastric pain and with the same capsaicin dose they reported greater pain intensity than controls.23 In addition, higher pain intensity at pain threshold stimulation and greater sustained levels of pain after reaching the target of moderate pain were seen in FD.
Capsaicin stimulation was chosen as the experimental stimulus because of the extensive physiological implication of the multimodal TRPV1 pathways not only in gastrointestinal sensation and nociception but also in the regulation of gastrointestinal motility, perfusion, secretion, mucosal homeostasis, and inflammation.22 Activation of the TRPV1 receptor includes some of the current etiological hypotheses for FD, such as intestinal inflammation or food hypersensitivity. Evidence for a role of TRPV1 pathways in FD was demonstrated in a receptor gene polymorphism (G315C) modulating upper gastrointestinal sensation in Japanese FD patients.56 In addition, an increased expression of TRPV1 and hypersensitivity to capsaicin have been demonstrated in IBS and other visceral pain syndromes.24,26,30,57
Our data extend previous results where fixed doses of 0.75 mg of oral capsaicin showed greater aggregate symptom scores in 54% of patients with FD than in healthy controls, with a large variability in responses to the fixed dose and a very low placebo response.27,58 Importantly, pain, the endpoint relevant for the assessment of sensitization, was not achieved in the majority of subjects.
Somatic sensitization was evaluated as evidence of central sensitization in FD. As a group, FD patients did not show somatic sensitization. Somatic sensation and pain thresholds and pain intensity during suprathreshold stimulation were similar in FD patients and controls. However, a substantial subgroup of between 20 and 60% of patients did show somatic hypersensitivity to hand or foot stimulation, depending on the stimulus intensity and location. The stimulation intensity at the pain tolerance threshold on the hand was negatively correlated with the rating of depression in FD patients, confirming the role of psychological factors such as depression in sensitization in FD.59 Although visceral hypersensitivity was not always accompanied by somatic hypersensitivity, most patients with somatic hypersensitivity also had visceral sensitization. It seems plausible that more intense visceral hypersensitivity may be driving the somatic sensitization. It is well-documented that tissue damage, such as a wound or inflammation, in one body part may lead to central sensitization.60–63 Somatic sensitization hypersensitivity was more common on the hand than on the foot. Currently discussed mechanisms for such differentiated cross-sensitization include an increased excitability of cross-innervating peripheral neurons or of spinal neurons receiving convergent viscero-somatic input and leading to central amplification of afferent information.64 A further explanatory mechanism for the difference between upper and lower extremity sensitization could be a selective hypervigilance in patients with an upper body symptom focus, as demonstrated in the selective response to a perceived threat.65 These data demonstrate a difference to what is known in IBS, where somatic hypersensitivity was similar using hand and foot stimulation.1,66,67
With this blinded and controlled capsaicin test, 97% of FD patients successfully reached the target of moderate abdominal pain, and the expectation-induced nocebo effect, defined as pain perception with application of an inert substance, was similarly low in patients and controls. In the small group of nocebo responders, the onset to pain was considerably delayed following nocebo (22 min) compared to capsaicin (8 min). This difference in response time indicates the different, expectation-derived analgesic process. These data together with the recently published good reproducibility of the test indicate its potential usefulness in the wider study of upper GI sensory changes.33 Further validation in a larger patient group in longitudinal studies is necessary to evaluate the test as a biomarker in FD.
Potential limitations of this study include: (1) the absence of detailed psychological profiling in our subjects, as these factors are known to affect sensory perception; (2) the absence of prospective subgrouping of patients according to the Rome III criteria The validated Leed’s Dyspepsia Questionnaire was used for symptom assessment, but this does not allow conclusive grouping according to the Rome criteria. There is ongoing discussion concerning the usefulness of the Rome III FD subgroups;68–71 (3) the exact site of capsaicin stimulation using the capsules was not proven. Whether the capsaicin acted primarily in the stomach or duodenum remains unsubstantiated. However, with the present understanding of the pathophysiology of FD, this may not be of prime importance, as both gastric and duodenal mechanisms are probably involved.33
In conclusion, a majority of patients with FD demonstrated visceral hypersensitivity to stimulation with capsaicin. Transient receptor potential vanilloid-1 pathways appear to be involved in visceral hypersensitivity in FD. The titration methodology used is simple and well-tolerated, with a low nocebo rate and a previously demonstrated high reproducibility. Somatic hypersensitivity was not evident in FD as a group, but it was present in a large subgroup of patients, especially in those with greater visceral sensitization. This implies that central sensitization may play a role in a segment of patients with FD. Other mechanisms, such as the well-documented accommodation and motility disorders, may be more relevant in other patients with this heterogeneous disorder.6,7,72 Further research into mechanisms driving the central sensitization, both peripheral and central, are now of considerable interest.
We wish to thank Ms. Tung Siew Lai, the research assistant for this study, for her expert help in study coordinating and subjects recruitment. We also wish to thank Ms. Siang Nee Teoh and colleagues for their expert help in manufacture and randomization of the study capsules.
This study was funded in full by Singapore National Medical Research Council, Individual Research Grant. Grant number: NMRC/1120/2007.
No competing interests declared by any of the authors.
XL was involved in acquisition , analysis, and interpretation of data, administrative, technical, or material support, manuscript revision and submission; YC was involved in acquisition , analysis, and interpretation of data, drafting of the manuscript, statistical analysis, administrative, technical, or material support; RW contributed toward study supervision, administrative support, critical revision of the manuscript; KYH obtained funding, assisted in study supervision, critical revision of manuscript for important intellectual content; CWS was involved in study concept and design, obtained funding, study supervision, drafting and critical revision of the manuscript, analysis and interpretation of data.