Abstract
Objective: Addiction is a chronic relapsing brain disease. Brain structural abnormalities may constitute an abnormal neural network that underlies the risk of drug dependence. We hypothesized that individuals with Betel Quid Dependence (BQD) have functional connectivity alterations that can be described by long- and short-range functional connectivity density(FCD) maps. Methods: We tested this hypothesis using functional magnetic resonance imaging (fMRI) data from subjects of the Han ethnic group in Hainan, China. Here, we examined BQD individuals (n = 33) and age-, sex-, and education-matched healthy controls (HCs) (n = 32) in a rs-fMRI study to observe FCD alterations associated with the severity of BQD. Results: Compared with HCs, long-range FCD was decreased in the right anterior cingulate cortex (ACC) and increased in the left cerebellum posterior lobe (CPL) and bilateral inferior parietal lobule (IPL) in the BQD group. Short-range FCD was reduced in the right ACC and left dorsolateral prefrontal cortex (dlPFC), and increased in the left CPL. The short-range FCD alteration in the right ACC displayed a negative correlation with the Betel Quid Dependence Scale (BQDS) (r=-0.432, P=0.012), and the long-range FCD alteration of left IPL showed a positive correlation with the duration of BQD(r=0.519, P=0.002) in BQD individuals. Conclusions: fMRI revealed differences in long- and short- range FCD in BQD individuals, and these alterations might be due to BQ chewing, BQ dependency, or risk factors for developing BQD.
Introduction
Areca nut(AN) can be chewed by wrapping in a betel leaf or with tobacco (betel quid, BQ), and its composition differs among diverse populations and areas [1]. For instance, BQ in Hainan province of China is a combination of fresh areca nut and hydrated lime, such as aqueous calcium hydroxide paste, wrapped in betel leaves without tobacco or other ingredients. BQ is broadly used by people of all ages around the globe, particularly in southeast Asia. The alkaloid arecoline contains parasympathomimetic properties that stimulate the muscarinic and nicotinic receptors, and therefore AN primarily affects the central and the autonomic nervous systems [2]. Frequent consumers indicate euphoria(a sense of warmth), happiness, salivation, elevated alertness, anti-migraine effects, palpitation and increased working capabilities [3]. BQ consumption is related to a dependency syndrome that consists of mild euphoria, enhanced concentration, postprandial satisfaction, and relaxation and a withdrawal syndrome related to mood swings, insomnia, anxiety and irritability. The severity of BQ dependence, is similar to that of amphetamine [4]. The first instrument to measure BQ dependence, the Betel Quid Dependence Scale (BQDS), was recently developed and initially validated by Lee [5].
BQ ranks among the most broadly consumed psychoactive substances worldwide following only nicotine, ethanol and caffeine, and its consumer population accounts for almost 10% of the global population. From a perspective of public health, BQ chewing is an important behavior because it is associated with various health problems, most notably oral cavity cancer and many precancerous lesions including leukoplakia and oral submucous fibrosis [6]. The International Agency for Research on Cancer review concluded that AN can be carcinogenic to people and may lead to cancers of the pharynx, oral cavity, esophagus, uterus, liver, and biliary tracts [7].
However, most studies regarding the behavior of chewing BQ have been confined to either epidemiological or biological investigations [5,8]. Researchers have conducted limited research aimed at determining the psychological and behavioral factors that cause people to start and/or maintain BQ consumption. Why people develop a BQ chewing habit remains unknown. Neuroimaging is a useful tool for investigating the drug-dependent brain. Brain structural abnormalities may constitute an abnormal neural network that might underlie the risk of drug dependence [9]. Our previous report was the first to use voxel-based morphometry (VBM) to disclose an association between BQ addiction behavior and gray matter (GM) volume[10]. Compared with the control group, VBM revealed that the BQD individuals had significantly decreased GM volume in the mid-brain, right rostral anterior cingulate cortex (ACC), bilateral dorsolateral prefrontal cortex (dlPFC) and right superior temporal gyrus (rSTG), while their GM volumes were increased in the right precuneus and right hippocampal. The GM volumes of the right rostral ACC and left dlPFC and midbrain volumes were negatively correlated with the duration of BQD and BQDS, respectively [10]. These results suggest that the neurobiological basis of BQD may be dysfunction of the emotion system, cognitive system, and reward system. However, the effects of BQD on neural activity remain largely unknown. Previous studies have indicated decreased activation of dlPFC combined with increased activity in the ventral striatum (VS) in methamphetamine-dependent patients compared with healthy controls using a risky decision-making task[11]. In terms of neural activation, task-associated changes only involve a small percentage (approximately 5%) of the total sum of cerebral activity [12]. Intrinsic activity expends greater energy of the brain compared with external stimuli [13]. Therefore, it is necessary to determine how the brain distributes resources to comprehend the neural mechanisms related to BQD, which remains unclear.
The connections between spontaneous fluctuations of blood oxygen level-dependent (BOLD) signals in diverse areas of brain in the “resting” state represent the resting-state functional connectivity, which is considered one type of functional organization of the brain [14]. Since functional connectivity is more readily applicable in a clinical environment, such as Alzheimer's disease or depression, than functional activation MRI, some research groups have begun analyzing the functional connectivity in the brain in the resting-state in various neuropsychiatric diseases [15,16]. In addition, as frequently used methods of brain functional connectivity analysis, neither independent component analysis (ICA) nor region of interest analysis (ROI) can examine multiple networks and measure the joint strength between brain regions [17]. Moreover, the ROI-based method is constrained by prior hypothesis. To overcome these shortcomings, analysis based on graph theory, such as functional connectivity density (FCD) mapping, is recommended [18] for large-scale, bias-free, and prior-hypothesis-free whole brain network analysis. FCD mapping can be derived from degree centrality calculations by calculating the degree centrality of each voxel in the whole-brain cortex and subcortical area [19,20,21]. However, no research has been carried out to study the FCD alteration in BQD.
We hypothesized that BQD patients are subject to alterations of functional connectivity that can be described by long- and short-range FCD maps. Therefore, in this research, we sought to produce voxel-wise long- and short-range FCD maps by calculating the brain network degree, to determine the extent of alteration of functional connectivity in BQD patients.
Materials and Methods
Ethics statement
The present study was approved by the research ethics review board of Hainan General Hospital, Haikou, China, in accordance with the Declaration of Helsinki (2000). A letter of consent was read and signed by each subject before inclusion in this study.
Inclusion and exclusion criteria
The study subjects were native people consuming BQ and exclusively located in Wanning City in Hainan province. The criteria for BQD group inclusion were as follows: (1) age between 18 and 60 years; (2) usage of BQ without tobacco for more than 5 years to exclude the influence of nicotine; (3)BQ chewer with BQD diagnosed according to BQD Scale (BQDS)>4, self-rating anxiety scale (SAS)<50 and self-rating depression scale(SDS)<50; (4) not taking antidepressant drugs or psychotropic drugs; (5) without self-claimed systemic illnesses(e.g. diabetes mellitus, cardiovascular disease, neurological disorder, thyroid disorders, renal disorders and epilepsy); (6) no present or recent history of receiving any Axis I psychiatric and/or substance diseases; (7) no contraindication to MRI examination; (8) right-handed; and (9) ability to read and write Chinese.
The inclusion criteria for control group were as follows: (1) age between 18 and 60 years; (2) without using all forms of BQ, areca nut and tobacco; (3) not taking antidepressant drugs or psychotropic drugs; (4) without self-claimed systemic illnesses(e.g. diabetes mellitus, cardiovascular disease, neurological disorder, thyroid disorders, renal disorders and epilepsy); (5) no present or recent history of receiving any Axis I psychiatric and/or substance diseases; (6) no contraindication to MRI examination; (7) right-handed; and (8) ability to read and write Chinese.
Study participants
At first, 36 control individuals and 38 BQD volunteers recruited from Wanning City of Hainan province underwent MRI scanning, but only 32 control individuals and 33 BQD volunteers were included because of angiocavernoma, lacunar infarction, arachnoid cyst and excessive head movement respectively.
Questionnaire
We distributed questionnaires to obtain information on age, sex, educational status, monthly income, daily dosage of BQ, duration of BQ chewing habit and time span of quid placement in the oral cavity in simple Chinese to all participants (n=65). Since liquor plays a significant role in Chinese society, we inquired about the intake of alcohol during the past 30 days, including the number of drinks per occasion and average frequency of drinking. Alcohol consumption was recorded in the form of original data according to the individuals' statements. However, most individuals were not able to recall the exact quantity of daily consumption. Instead, they could only provide descriptions as vague as “one pack” or “half a pack.” So we had to convert such descriptions into numbers, based on our knowledge of the commercial packs of each product. For examples, the alcohol is sold in small bottles (100 mL) and standard bottles (500 mL). Alcoholic beverages were recorded as beer or white spirit (a Chinese distilled beverage with alcohol content at about 50%), which were substantially consumed in Mainland China. One gram pure ethanol would approximately correspond to 18.3 cc of beer, or to 2 cc of white spirit.
MRI Data Acquisition
In the Department of Radiology, Hainan General Hospital, MR imaging data were obtained using a 3.0 T MRI scanner (Magnetom Verio, Siemens, Erlangen, Germany) with a standard 6-channel head coil. During the MR scanning, participants were required to remain awake with their eyes closed and heads still without thinking about a specific subject. First, the BOLD contrast of whole-brain functional images was acquired by a T2*-weighted EPI sequence (repetition time= 2000 ms, echo time=30 ms, field of view= 240×240 mm2, flip angle = 80°, image matrix = 64×64, voxel size= 3.75 ×3.75×5 mm3, without gap, each functional run comprised 240 volumes, and each brain volume included 31 axial slices). Then, a high-resolution T1-weighted structural image was obtained in the sagittal orientation using a magnetization prepared rapid acquisition gradient echo (MPRAGE) sequence (repetition time=2300 ms, echo time=2.9 ms, TI=900 ms, field of view=256×256 mm2, flip angle=9°, in-plane matrix= 256×256, slice thickness = 1mm, no gap, and voxel dimension=1×1×1.33mm3).
MRI Data Preprocessing
The toolbox Data Processing Assistant for Resting-State functional MR imaging (DPARSF; http://www.restfmri.net/forum/DPARSF)[22] was used for preprocessing fMRI imaging data using statistical parametric mapping (SPM8; http://www.fil.ion.ucl.ac.uk/spm/) and the rs-fMRI data analysis toolkit (REST1.8; http://www.restfmri.net). The first 10 volumes of each functional time series were removed to ensure an equilibrium state of magnetization. Slice timing and realignment for head motion correction were performed. Spatial normalization to the standard Montreal Neurological Institute(MNI) echo-planar imaging template was performed in the Statistical Parametric Mapping package; the functional images were then spatially normalized to standard coordinates and resampled to 3 × 3 × 3 mm3[23]. All subjects with more than 1.5mm translation head motion or more than 1.5° rotation in all directions were excluded. Finally, we smoothed the resampled images with a Gaussian kernel of 4 mm. We then performed linear trend and band-pass filtering (0.01-0.08Hz) [24,25] to remove the influence of low-frequency drift and high-frequency noise.
Long- and short-range FCD calculation
Based on Pearson correlations between the time course of a given voxel and that of other voxels, the number of functional connections of a given voxel was considered a degree of a node in a binary graph [20]. The detailed procedure of the computing of local and global FCD is given in[26]. A neighborhood strategy was adopted to determine long- and short-range FCD, in which two main points were considered [19]. First, we defined functional connectivity in the whole brain between a given voxel with every other voxel, with a correlation threshold of r > 0.25[20]. Second, the long- and short- range FCD were defined on the basis of the neighborhood strategy, which means that we defined voxels that met a correlation threshold of r > 0.25 inside their neighborhood (radius sphere ≤12 mm) as short-range FCD and outside their neighborhood (radius sphere >12 mm) as long-range FCD. For further data analysis, we converted the long- and short-range FCD maps to Z scores [20], and spatially smoothed them with an 8-mm full-width at half-maximum Gaussian kernel using SPM8.
Statistical analysis
Using SPSS software (version 16.0; SPSS, Inc., Chicago, IL), we compared the demographic and clinical data between the two groups. An independent two-sample t test was used to compare the age, years of education, monthly income, SDS, SAS and alcohol in the last 30 days, and the Chi-square test was used to compare gender difference and handedness. A P value<0.01 was regarded as statistically significant.
SPM8 was used for rs-fMRI data analysis. We performed a second-level random-effect one-sample t test to display the spatial distribution of short- and long- range FCD in healthy controls (HCs) and the BQD group, respectively. Then, we carried out second-level random-effect two-sample t tests to compare BQD individuals and HCs over their short- and long-range FCD. Using the AlphaSim program with REST software (http://www.restfmri.net), we conducted a multiple comparison correction in which the significance level was set at an uncorrected P<0.01, with a cluster size>65 mm3, corresponding to a corrected P<0.05 (for individual voxels, P=0.01, FWHM=8 mm, r=5mm, iterations=1000). The long-/short-range FCD changes (displayed as the mean ΔFCD) in brain regions with significant differences in FCD between HCs and the BQD group were extracted from each patient using spherical ROIs (r=6 mm)[27], and each ROI was centered at the point of the peak t value.
The correlations between the long-/short-range FCD in these brain regions and the variation of the corresponding clinical data were evaluated using Pearson's correlation analysis with SPSS for Windows (version 16.0, Chicago, IL). The significance level was set at P<0.05.
Results
Demographics and clinical characteristics
In the final data analysis, we included 65 subjects (33 BQD patients and 32 HC). The BQD individuals and control-group members exhibited no differences in sex or age. The results for emotional status evaluated by SAS and SDS failed to reach the cut-off for clinical significance on average, although the SDS scores of the BQD group were significantly lower than those of the controls. No significant differences were observed in education levels, monthly income, and intake of alcohol in the last 30 days between the two groups (P values >0.05). Individuals indicated that they had been chewing BQ with dependency syndrome for a mean duration of 20.6±6.9 years (range 7 to 31 years), a mean BQDS of 10±3.4 (range 5 to 16) and consumed an average of 342±106g/day BQ (range 200 to 500 g/day). BQD chewers placed BQ in their mouth for an average of 7.6±2.4 minutes (range 3 to 12) before spitting-out the remnants. The results were described in great detail in our VBM investigation report [10]. Table 1 depicts the demographics of the BQD and healthy control subjects.
Demographics and clinical characteristics of the participants. Note: Unless otherwise indicated, data are means ± standard deviations. N/A = not applicable, SAS=Self-Rating Anxiety Scale, SDS=Self-Rating Depression Scale. Values are expressed as the mean ± SD. The P values for gender distribution and handedness in the two groups were obtained by the chi-square test. The P values for differences in age, education, alcohol last 30 days, SAS, and SDS between the two patient groups were obtained by independent-samples t test

Rs-fMRI data
The results from the one-sample t test showed similar spatial distributions of long-range and short-range FCD in the HC and BQD group respectively (Fig. 1)(P<0.05, AlphaSim corrected). Short-range FCD was mainly distributed in the bilateral posterior cingulate cortex/precuneus (PCC/PCu), posterior parietal cortices, occipital lobe and prefrontal cortices (PFC), whereas long-range FCD was preferentially distributed in the bilateral PCC/PCu, cuneus, inferior parietal lobule (IPL), and PFC. In addition, either long-range or short-range FCD was distributed in brain areas belonging to the default mode network (DMN)[21], including the medial prefrontal cortex (MPFC), PCC/PCu, and IPL.
Spatial distribution of long- and short-range FCD differences within the BQD and HC groups respectively by one-sample t test. Either long-range or short-range FCD distributed in brain areas belonging to default mode network (DMN), including medial prefrontal cortex (MPFC), PCC/PCu, and IPL.
Spatial distribution of long- and short-range FCD differences within the BQD and HC groups respectively by one-sample t test. Either long-range or short-range FCD distributed in brain areas belonging to default mode network (DMN), including medial prefrontal cortex (MPFC), PCC/PCu, and IPL.
The differences in Long-/short-range FCD between the BQD group and HCs are displayed in Table 2 and Fig. 2(P<0.05, AlphaSim corrected). When we compared BQD patients with HCs, the former displayed decreasedt long-range FCD in the right ACC and increased long-range FCD in the left cerebellum posterior lobe (CPL) and bilateral IPL. With respect to short-range FCD, we observed a reduction in the right ACC and left dlPFC and an increase in the left CPL.
Differences in short- and long-rangeFCD between the BQD patient and control groups. Comparisons were performed at P<0.05, corrected for multiple comparisons. Abbreviations: L: left; R: right; ACC= anterior cingulate cortex; dlPFC= dorsolateral prefrontal cortex; CPL=cerebellum posterior lobe; IPL=inferior parietal lobule; BA: Brodmann's area. t: statistical value of peak voxel showing significant long- and short-range FCD differences between the two groups (negative values: BQD<HCs; positive values: BQD>H-Cs). MNI: Montreal Neurological Institute Coordinate System or Template; X, Y, Z: coordinates of primary peak locations in the MNI space

Two-sample t test and paired t test results for long- and short-range FCD alterations between the BQD and HC groups. The results show decreased long-range FCD mainly in the right anterior cingulate cortex (ACC) and increased long-range FCD in the bilateral inferior parietal lobule (IPL) and left cerebellum posterior lobe (CPL). The results show decreased short-range FCD mainly in the right ACC and left dorsolateral prefrontal cortex (dlPFC) and increased short-range FCD in the left CPL.
Two-sample t test and paired t test results for long- and short-range FCD alterations between the BQD and HC groups. The results show decreased long-range FCD mainly in the right anterior cingulate cortex (ACC) and increased long-range FCD in the bilateral inferior parietal lobule (IPL) and left cerebellum posterior lobe (CPL). The results show decreased short-range FCD mainly in the right ACC and left dorsolateral prefrontal cortex (dlPFC) and increased short-range FCD in the left CPL.
Correlation Analysis
As shown in Fig. 3, as revealed by Pearson's correlation analyses, short-range FCD alteration of the right ACC displayed a negative correlation with BQDS(r=-0.432, P=0.012), and long-range FCD alteration of left IPL showed a positive correlation with duration of BQD(r=0.519, P=0.002) in BQD individuals. Our results showed no correlation between the SAS and above-mentioned brain areas with long-/short-range FCD alterations. In addition, there were no correlations between SDS and neural activity changes, although the SDS scores in the BQD group had a downward trend.
Correlation results between FCD alteration and duration and BQDS. Pearson's correlation reveals that short-range FCD alteration in the right ACC showed a negative correlation with BQDS (r=-0.432, P=0.012) and that long-range FCD alteration in the left IPL showed a positive correlation with duration of BQD (r=0.519, P=0.002) in BQD individuals. (BQD = Betel quid dependence; BQDS = Betel quid dependence score; ACC= anterior cingulate cortex; IPL= inferior parietal lobule).
Correlation results between FCD alteration and duration and BQDS. Pearson's correlation reveals that short-range FCD alteration in the right ACC showed a negative correlation with BQDS (r=-0.432, P=0.012) and that long-range FCD alteration in the left IPL showed a positive correlation with duration of BQD (r=0.519, P=0.002) in BQD individuals. (BQD = Betel quid dependence; BQDS = Betel quid dependence score; ACC= anterior cingulate cortex; IPL= inferior parietal lobule).
Discussion
In this study, BQD patients exhibited decreased long- and short-range FCD mostly in the right ACC but increased FCD in the left CPL compared with healthy controls. In addition, we found that, in BQD patients, long-range FCD was further increased in the bilateral IPL, whereas short-range FCD was decreased in the left dIPFC. We also found that BQDS was negatively correlated with short-range FCD in the right ACC and duration was positively correlated with long-range FCD in the left IPL. These variations of long- and short-range FCD might be due to BQ chewing, BQ dependency, or risk factors for developing BQD. This is the first research to assess long- and short-range FCD alterations in BQD individuals.
There are interactions between local (short-range) and distant (long-range) information processing in the “resting” human brain [21]; these interactions could optimize brain information processing to increase efficiency and reduce costs. Short- and long- range FCD represent projections from the proximal area and remote brain systems, respectively, and thus might facilitate the elucidation of brain information intermingled with projections from both local and distant areas. Our findings revealed changes in long- and short-range FCD in patients with BQD compared with normal controls, and these changes mainly involved some components of the DMN (ACC and IPL) and frontal (dIPFC) and cerebellar (CPL) lobes. These discoveries might to some extent serve as evidence of the reorganization of the brain network in BQD individuals.
We might consider the DMN the most significant component of the “resting” brain network [28] in addition to a set of brain regions that are deactivated during demanding cognitive tasks and anti-correlated with the fronto-parietal regions [29,30,31,32]. The function of the DMN is largely connected with information processing that is self-relevant and internally directed [31]. The DMN comprises brain regions [21] that are the main location of the overlap of long- and short-range FCD in the resting state, further emphasizing the role of the DMN in processing both adjacent and remote information in the brain. The right ACC and bilateral IPL, which belong to the DMN, had abnormal brain activity in the BQD group. The role of the ACC has led to its inclusion in many major theories of addiction, in which it is believed to form part of an inhibitory system that exercises control over reward-related behavior [33,34]. Studies of heroin and cocaine-dependent patients have shown that hypoactivation in the rostral ACC is connected to deficiencies in response inhibition and impulse control [35,36]. The influence of the ACC dysfunction in addicts was also confirmed by the model of addiction proposed by Volkow et al. [37,38]. In the present study, short-range FCD in the right ACC was also negatively correlated with BQDS, we argued that BQD can cause damage to the brain structure and its function, progressively. Based on this correlation, we speculate that abnormal neural activity in the right ACC might play an indispensable role in BQ chewing, BQ dependency, or susceptibility to developing BQD.
We also observed that increased long-range FCD in the bilateral IPL. There are an associations between the IPL and verbal fluency, verbal working memory, skill learning and acquisition, as well as complex sequential motor behavior [39,40]. Verbal fluency deficiency has been demonstrated in cocaine-dependent individuals [41], and motor dysfunction has been observed in methamphetamine-dependent patients. Although neuropsychological evaluations were not performed in our study, we speculate that BQD impairs distant information processing in the bilateral IPL. Overall, this study reveals impairment of components of the DMN (ACC, IPL) in BQD individuals from a perspective of functional connectivity.
Additionally, we observed the decreased of short-range FCD in the left dlPFC. Various neurocognitive studies have indicated a correlation between cognitive management and a special cortico-subcortical network that includes the rostral ACC and dlPFC [42]. The dorsal ACC and the posterior part of the left dlPFC (BA6/8) have been well documented as the important areas that make contributions to top-down attention control and performance adjustment based on contextual demands [43]. The left dlPFC imposes a top-down attentional set and biasing prior to slate-stage selection by the dorsal ACC [44]. When implementing top-down attentional control, the dorsal ACC must strengthen its activity to compensate for the low activity of the left dlPFC [45]. At least to some degree, the reduced long- and short-range FCD mainly in the right rostral ACC and the decrease in short-range FCD in the left dlPFC may be correlated with cognitive management and behavioral dysfunctions directed by a goal in BQ addiction [46,47].
Interestingly, we found increased long- and short-range FCD in the left CPL in the BDQ group. Previous PET and fMRI studies have revealed that drug-conditioned cues elicit increased metabolism and activation of the cerebellum [48,49].Glucose metabolism is greatly increased in the cerebellum when addicts perform reward expectation tasks [50], which suggests that the cerebellum is also included in drug-conditioned memories in addicts. In addition, the cerebellum has been described to play a compensatory role in inhibitory control [51] and decision-making behavior in addicts [52]. Based on these previous studies and our results above, we argue that altered long- and short-range FCD in the left CPL might reflect neuroadaptation and reorganization of the cerebellar functional network caused by BQD. Further research is needed to validate this conclusion.
Limitations
There are some limitations to this research. First, there is no neuropsychological tests to define BQD patients, which restricted our understanding of the outcomes. Second, a longitudinal study is needed to substantiate the findings of the development of BQD in the same subject group instead of in different groups.
Conclusion
In conclusion, prominent FCD alteration was found mainly in brain areas related to the DMN (right rostral ACC/bilateral IPL), reward circuit (right ACC/dlPFC) and cerebellum, which might be due to BQ chewing, BQ dependency, or risk factors for developing BQD. Our findings need to be further verified.
Acknowledgements
The authors extend heartfelt gratitude to all of the volunteers for their participation in this study, and to the staff at the Department of Radiology of Hainan General Hospital for their dedicated support.
Funding
This study was financed by grants from the Natural Science Foundation of China (Grant No. 81260218 to Jianjun Li, Grant No. 81460261 to Feng Chen), the Natural Science Foundation of Hainan Province (Grant No. 813201 to Tao Liu), the key science and technology project of Hainan Province (Grant No. ZDXM20120047 to Jianjun Li), the Social Science development Foundation of Hainan province (Grant No. SF201312 to Feng Chen, Grant No. SF201414 to Tao Liu), the Hainan Health Institution Project (Grant No. 2012PT-06 to Tao Liu) and the National clinical key subject construction project.
Disclosure Statement
The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.
References
T. Liu and J. Li contributed equally to this work.