CLUSTER ANALYSIS OF THE PATHOGENETIC RELATIONSHIPS OF METABOLIC PARAMETERS IN PATIENTS WITH NON-ALCOHOLIC FATTY LIVER DISEASE ON THE BACKGROUND OF HYPERTENSION Текст научной статьи по специальности «Клиническая медицина»

MEDICINE

CLUSTER ANALYSIS OF THE PATHOGENETIC RELATIONSHIPS OF METABOLIC PARAMETERS IN PATIENTS WITH NON-ALCOHOLIC FATTY LIVER DISEASE ON THE BACKGROUND OF HYPERTENSION

Professor Oleg Babak, PhD student Anna Bashkirova,

Kharkiv, Ukraine, Kharkv National Medical University DOI: https://doi.org/10.31435/rsglobal_ws/31102019/6717

ARTICLE INFO

ABSTRACT

Received: 15 August 2019 Accepted: 20 October 2019 Published: 31 October 2019

KEYWORDS

NAFLD, hypertension, endothelial lipase, cluster analysis.

The aim of the study was to conduct a cluster analysis of pathogenetic relationships between metabolic parameters, endothelial lipase levels, the severity of steatosis, and clinical parameters in patients with non-alcoholic fatty liver disease with hypertension. To analyze pathogenetic relationships, a cluster analysis was performed with the distribution of parameters into 4 clusters using the Ward's method. The most dense metabolic link by cluster analysis endothelial lipase forms with NAFLD liver fat score (2.639 cu), HbAlC (2.084 cu), total cholesterol (2.272 cu), and alcohol units (2.797 cu).

Citation: Oleg Babak, Anna Bashkirova. (2019) Cluster Analysis of the Pathogenetic Relationships of Metabolic Parameters in Patients with Non-Alcoholic Fatty Liver Disease on the Background of Hypertension. World Science. 10(50), Vol.1. doi: 10.31435/rsglobal_ws/31102019/6717

Copyright: © 2019 Oleg Babak, Anna Bashkirova. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

Non-alcoholic fatty liver disease (NAFLD) is the most common cause of impaired liver function in adults and children [1]. NAFLD covers the histological spectrum from simple steatosis to nonalcoholic steatohepatitis (NASH), progressive fibrosis and cirrhosis [2]. Simple steatosis without fibrosis or inflammation in most cases has a benign clinical course, but often leads to an increase in mortality [3]. The possible role of NAFLD as a risk factor for the development of cardiovascular diseases has been discussed for a long time, and only recent data have demonstrated the existing relationship between these conditions [4]. Insulin resistance is often detected in patients with NAFLD, as in patients without obesity and diabetes [5]. NAFLD is often associated with components of the metabolic syndrome, such as type 2 diabetes mellitus (T2DM), obesity, hypertension, and dyslipidemia [7]. However, an increasing number of patients with a normal body mass index (BMI) have been described, with central obesity and latent insulin resistance. [6] Several studies have shown that adopting a healthy lifestyle, reducing weight, and proactively correcting individual components of the metabolic syndrome can help prevent, slow down, or reverse liver damage associated with NAFLD [8].

Endothelial lipase (EL) is a strong determinant of the structural and functional properties of high density lipoproteins (HDL) [9]. EL is a new marker of cardiovascular risk, which is closely associated with dyslipidemia and insulin resistance and has hardly been studied in the presence of NAFLD [10].

Regardless of this, NAFLD increases the risk of premature cardiovascular disease and related mortality, therefore, research and monitoring of the metabolic function of the liver and early detection of accumulation of EL, as well as the relationships between them, are of great importance.

The purpose of the study was to conduct a cluster analysis of pathogenetic relationships between metabolic parameters, EL levels and clinical parameters in patients with liver steatosis on the background of hypertension.

Materials and methods 80 patients have been examined on the basis department of internal medicine №1 of Kharkiv National Medical University and National Institute of Therapy named by L.T. Malaya of National Academy of Medical Sciences of Ukraine. The patients have been divided into three groups according to the severity of liver steatosis. The first group consisted of 16 patients with hypertension without laboratory or instrumental signs of liver steatosis (hypertension group). Patients who, in addition to hypertension, had signs of steatosis during ultrasound and normal level of transaminases (ALT, AST), formed a group with moderate liver steatosis (MLS, n = 20). Patients with hypertension who, in addition to the echoscopic features of hepatic steatosis had increased level of transaminases, were assigned to the group with severe liver steatosis (group SLS, n = 24). The control group consisted of 20 practically healthy individuals. The patients' ages ranged from 45 to 60 years, with an average age of 52.12 + 5.24 years. Among them 28 were female (46.66%) and 32 were male (53.33%).

for identification of liver steatosis and its severity we have used liver fat index (NAFLD liver fat score), which includes such indicators as the presence of metabolic syndrome and T2DM, serum insulin level, AST and the ratio AST/ALT and is calculated by the formula [11]:

NAFLD liver fat score= - 2.89+1.18xmetabolic syndrome (yes=1/no=0)+ 0.45xtype 2 diabetes (yes=2/no=0)+0.15x fasting serum Insulin (mU/L)+ 0.04 x fasting serum AST(U/L) - 0.94 x AST/ALT.

The FIB-4 index has been used to identify liver fibrosis, which includes indicators such as AST, ALT, platelet count, and is calculated by the formula [12]:

FIB4 = Age (years) xAST (IU/L)/platelet count (x109/L)xVALT (IU/L) Serum endothelial lipase (EL) concentration was determined by enzyme-linked immunosorbent assay using Aviscera Bioscience INC reagent kit (USA) using a Labline 90 enzyme immunoassay analyser.

For excluding the alcoholic genesis of NAFLD all patients have been interviewed to determine alcohol units. This test has international standardization and allows detecting alcohol abuse by the formula: Alcohol units = amount (liters) x alcoholic strength (%) x 0.789 Alcohol abuse was eliminated by less than 14 units per week regardless of gender [13]. In order to monitor the implementation of dietary recommendations, we have used a questionnaire designed by the original questionnaire, which asked patients about the consumption of 15 basic foods that are not recommended for overweight, carbohydrate metabolism disorders and liver steatosis.

The statistical processing of the survey data has been performed using Microsoft Exel and Statistica 7.0 using standard methods of virion statistics.

Results and discussion. Results of studies are presented in table 1.

Table 1. Anthrometric, laboratory and surrogate ratios indicating the severity of liver steatosis

Parameter Control, n=20 Hypertension group, n=16 MLS group, n=20 SLS group, n=24 Significance of difference, P

0 1 2 3

Mean SD Mean SD Mean SD Mean SD

1 2 3 4 5 6 7 8 9 10

AST/ALT, U 0.73 0.34 0,95 0,26 1,19 0,38 1,75 0,77 12, 23, 13

ALT, U/L 18,15 7,26 22,44 6,29 24,65 7,19 62,54 40,78 23, 13

AST, U/L 11,45 3,78 24,56 7,84 22,90 10,18 38,71 20,70 23, 13

AP, mmol/l 1334,90 464,11 1422,50 302,46 1457,13 384,29 1834,06 690,13 13, 23

NAFLD liver fat score -1,93 0,65 -0,308 1,14 2,308 2,43 4,48 3,21 For all groups < 0,001

Fib-4 0,43 0,16 1,07 0,36 1,14 0,72 1,36 0,63 With control - all groups< 0.0001 13, 23

BMI, kg/m2 21,44 1,57 25,91 3,42 30,00 2,79 29,04 5,44 01, 02, 03, 12

WC, cm 75,50 6,83 79,31 8,58 98,08 10,53 104,10 8,67 02, 03, 12, 13, 23

WC/height, U 0,44 0,03 0,47 0,04 0,57 0,05 0,60 0,04 01, 02, 03, 12, 13, 23

Continuation o ? table

1 2 3 4 5 6 7 8 9 10

SBP, mm Hg 116,00 4,17 161,56 17,77 163,89 17,54 169,17 22,20 01, 02, 03

DBP, mm Hg 73,50 5,16 101,56 7,47 102,78 8,26 101,46 9,94 01, 02, 03

Cholesterol, mmol/l 3,85 0,77 5,25 1,47 5,74 0,85 5,80 1,42 01, 02, 03

Triglycerides, mmol/l 0,92 0,16 1,13 0,38 1,70 0,83 0,33 1,96 0,67 0,37 12, 13, 23

HDL, mmol/l 1,77 0,28 1,47 0,42 1,42 0,30 1,20 0,27 13, 23

LDL, mmol/l 2,36 0,46 3,45 1,41 3,34 0,85 3,75 1,25

VLDL, mmol/l 0,38 0,05 0,56 0,16 0,70 0,40 0,92 0,31 13, 23

EL, ng / ml 8,23 2,47 10,54 2,69 13,21 3,59 13,71 3,71 01, 02, 03, 12, 13

Diet 2,36 0,81 2,57 0,53 2,64 1,15 2,08 0,86

Alcohol units 4,26 2,27 4,29 1,82 6,39 2,99 6,62 2,98 02, 03, 12, 13

Fasting glucose, mmol/l 4,36 0,72 5,01 0,60 6,32 1,75 5,73 0,91 12, 13

Fasting insulin, mU/l 7, 91 3,71 17,77 6,86 24,51 9,49 33,28 13,82 12, 13, 23

HOMA-IR 1,55 0,85 3,61 1,80 7,02 4,76 8,35 5,25 12, 13

HbAlC, % - - 5,40 0,63 6,64 1,76 5,79 0,49 12, 23, 13

For the analysis of pathogenetic relationships, a cluster analysis was performed. The graph of parameter merging using Ward's method showed that it is advisable to distribute data into 3-4 clusters (Fig. 1).

Plot of Linkage Distances across Steps

Euclidean distances 3000 -----------

2500 ■

2000 ■

<u o

ro 1500 ■

v>

Q

<1J

jo 1000 ■ g

500 ■

0 ■ ■

-500 ---------------

0 2 4 6 8 10 12 14

- Linkage

Step Distance

Fig. 1. Cluster aggregation of data parameters

The first and second cluster illustrates the existence of offline hypertension and the close relationship of hyperinsulinism with BMI.

The third cluster covered liver steatosis associated with alcohol consumption, compensation for carbohydrate metabolism (by HbAlC) and the level of total cholesterol and EL.

The fourth cluster demonstrates the connection of lipid profile parameters with the patient's nutritional preferences (Table 2).

Table 2. Clustering of model components containing parameters of lipid-carbohydrate metabolism with hypertension and liver steatosis_

Cluster 1 Cluster 2 Cluster 3 Cluster 4

Parameter Mean Parameter Mean Parameter Mean Parameter Mean

SBP 33,04 Fasting insulin 8,520 HbA1C 2,084 Triglycerides 0,840

DBP 33,04 BMI 8,520 EL 5,386 HDL 1,231

Cholesterol 2,272 LDL 1,864

Alcohol units 2,797 VLDL 1,470

NAFLD liver fat score 2,639 Diet 1,306

WC/height 1,472

Grouping the results of the examination of patients allowed us to identify 4 main clusters. Cluster 1 - overweight patients with abdominal fat distribution and moderate hypertension, moderate hyperinsulinism with prediabetic levels of HbA1C, increased levels of total cholesterol, LDL and borderline HDL, moderate steatosis (table 3).

Table 3. Basic statistical analysis of clinical and laboratory parameters Cluster 1

Parameters Valid N Mean Minimum Maximum Variance Std.Dev. Coef.Var.

WC/height 13 0,5391 0,3593 0,6221 0,00544 0,073755 13,68212

Fasting insulin 13 25,0715 15,7000 46,0500 93,55990 9,672637 38,58015

HbA1C 13 6,1092 4,3300 10,3000 2,36916 1,539207 25,19477

EL 13 12,0465 7,0665 17,4200 9,18047 3,029928 25,15198

Cholesterol 13 5,9615 4,0200 7,7900 1,40600 1,185748 19,88996

Triglycerides 13 1,4846 0,7900 4,1600 0,73911 0,859715 57,90828

HDL 13 1,4046 0,6900 2,0000 0,13968 0,373734 26,60755

LDL 13 3,7492 2,2800 5,3500 0,86937 0,932402 24,86917

VLDL 13 0,6277 0,1900 1,8700 0,16587 0,407270 64,88377

SBP 13 142,6923 130,0000 150,0000 52,56410 7,250111 5,08094

DBP 13 94,6154 80,0000 100,0000 56,08974 7,489309 7,91553

BMI 13 27,5394 16,8525 32,4500 17,56233 4,190744 15,21728

Alcohol units 13 6,4231 2,0000 9,2000 5,46692 2,338145 36,40226

Diet 13 2,7692 1,0000 4,0000 0,85897 0,926809 33,46809

NAFLD liver fat score 13 2,1543 -0,8161 6,1263 3,98455 1,996134 92,65719

Cluster 2 included patients with grade 1 obesity with severe abdominal fat distribution and severe hypertension, a pre-diabetic level of HBA1C, an increase in total cholesterol, triglycerides, LDL and a distinct decrease in HDL and severe liver steatosis (Table 4).

Table 4. Basic statistical analysis of clinical and laboratory indicators of representatives of Cluster 2

Parameters Valid N Mean Minimum Maximum Variance Std.Dev. Coef.Var.

WC/height 11 0,5934 0,5058 0,7317 0,0038 0,06164 10,38688

Fasting insulin 11 53,0082 35,3300 73,3500 172,3410 13,12787 24,76574

HbA1C 11 6,0109 4,8600 7,4400 0,9214 0,95992 15,96965

EL 11 13,6655 7,7600 19,7200 21,8332 4,67261 34,19283

Cholesterol 11 6,4600 4,1500 8,5500 1,4985 1,22414 18,94954

Triglycerides 11 2,5227 0,9700 5,6900 1,8036 1,34298 53,23534

HDL 11 1,0718 0,7600 1,5600 0,0545 0,23340 21,77624

LDL 11 4,2218 1,9400 6,4000 1,8807 1,37138 32,48310

VLDL 11 1,0982 0,3700 2,5600 0,4095 0,63995 58,27362

SBP 11 175,9091 160,0000 180,0000 54,0909 7,35465 4,18094

DBP 11 105,0000 100,0000 120,0000 45,0000 6,70820 6,38877

BMI 11 30,0625 22,9481 48,3343 51,2784 7,16089 23,82002

Alcohol units 11 6,1909 1,5000 10,0000 8,6669 2,94396 47,55298

Diet 11 2,9091 1,0000 4,0000 1,0909 1,04447 35,90352

NAFLD liver fat score 11 6,8779 0,9228 10,9828 11,5463 3,39799 49,40435

Cluster 3 was composed of overweight patients with abdominal fat distribution, a slight increase in insulin at normal HbAlC, a slight increase in total cholesterol and LDL, normal triglycerides and marginal HDL (Table 5).

Table 5. Basic statistical analysis of clinical and laboratory indicators of representatives of Cluster 3

Parameters Valid N Mean Minimum Maximum Variance Std.Dev. Coef.Var.

WC/height 23 0,5448 0,4606 0,6875 0,00439 0,066239 12,1574

Fasting insulin 23 18,1643 4,8600 37,1000 85,67632 9,256150 50,9578

HbA1C 23 5,6678 4,3700 7,2800 0,61653 0,785192 13,8535

EL 23 11,2209 5,3300 18,9900 10,31109 3,211089 28,6171

Cholesterol 23 5,3009 2,8400 6,9900 1,65284 1,285627 24,2531

Triglycerides 23 1,3804 0,7500 2,6900 0,29569 0,543770 39,3912

HDL 22 1,3318 0,8000 2,3000 0,11832 0,343977 25,8276

LDL 22 3,2109 1,0000 5,1600 1,38308 1,176042 36,6265

VLDL 23 0,6991 0,3400 1,5000 0,10728 0,327538 46,8493

SBP 23 168,2609 150,0000 180,0000 53,65613 7,325034 4,3534

DBP 23 103,0435 100,0000 120,0000 31,22530 5,587960 5,4229

BMI 23 28,1174 22,6003 37,1255 13,17259 3,629407 12,9080

Alcohol units 23 5,3000 1,5000 11,5000 8,36909 2,892938 54,5837

Diet 23 3,0000 1,0000 4,0000 0,90909 0,953463 31,7821

NAFLD liver fat score 23 0,8359 -2,7277 4,9742 4,40681 2,099239 251,1388

Cluster 4 is the least numerically representative, but the most unexpected. It included patients with severe abdominal obesity, diabetic levels of HbAlC, moderate hyperinsulinism, with an increase in total cholesterol, TG, LDL and a decrease in HDL (Table 6).

Table 6. Basic statistical analysis of clinical and laboratory indicators of representatives of Cluster 4

Parameters Valid N Mean Minimum Maximum Variance Std.Dev. Coef.Var.

WC/height 5 0,5782 0,4821 0,6180 0,0032 0,05692 9,84298

Fasting insulin 5 21,9720 16,1700 27,5400 26,5516 5,15283 23,45179

HbA1C 5 6,6280 5,4200 9,9600 4,8687 2,20652 34,86912

EL 5 13,9380 9,2300 18,2800 14,5143 3,80977 27,33367

Cholesterol 5 5,6600 4,6000 6,7600 0,7388 0,85951 15,18561

Triglycerides 5 2,2500 1,2300 3,2900 0,8322 0,91222 40,54322

HDL 5 1,1440 0,7800 1,6100 0,1012 0,31817 27,81177

LDL 5 3,5060 2,4400 4,5000 0,6001 0,77468 22,09585

VLDL 5 1,0000 0,5000 1,4800 0,1811 0,42562 42,56172

SBP 5 203,0000 190,0000 240,0000 445,0000 21,09502 10,39164

DBP 5 114,0000 100,0000 130,0000 130,0000 11,40175 10,00154

BMI 5 30,1405 24,8016 33,9100 12,8248 3,58118 11,88161

Alcohol units 5 6,8800 4,5000 11,5000 10,6970 3,27063 47,53818

Diet 5 2,6000 2,0000 3,0000 0,3000 0,54772 21,06625

NAFLD liver fat score 5 1,7530 0,2018 2,9370 1,5633 1,25032 71,32395

For better visualization, we have reduced the average values of the parameters for the clusters into a common table (Table 7), which allows us to compare trends.

Table 7. The average values of clinica and laboratory indicators with the distribution of clusters

Parameters Cluster 1 Cluster 2 Cluster 3 Cluster 4

WC/height 0,54 0,59 0,54 0,58

Fasting insulin 25,07 53,01 18,16 21,97

HbA1C 6,11 6,01 5,67 6,33

EL 12,05 13,67 11,22 13,94

Cholesterol 5,96 6,46 5,30 5,66

Triglycerides 1,48 2,52 1,38 2,25

HDL 1,40 1,07 1,33 1,14

LDL 3,75 4,22 3,21 3,51

VLDL 0,63 1,10 0,70 1,00

SBP 142,69 175,91 168,26 203,00

DBP 94,62 105,00 103,04 114,00

BMI 27,54 30,06 28,12 30,14

Alcohol units 6,42 6,19 5,30 6,88

Diet 2,77 2,91 3,00 2,60

NAFLD liver fat score 2,15 6,88 0,84 1,75

The lowest variability of characteristics in the first cluster is inherent in indicators of blood pressure, BMI and insulin concentration, and the largest - in the severity of liver steatosis.

The lowest variability in the second cluster is inherent in blood pressure and anthropometric parameters, as well as indicators of carbohydrate metabolism. The variability of the severity of steatosis is half that of the previous group.

Cluster 3 from cluster 1 is distinguished by lower numbers of blood pressure, less severe liver steatosis and less alcohol abuse. The indicated group is determined by the relative stability of lipid

profile parameters, the stability of carbohydrate metabolism compensation, but the high variability of NAFLD. Thus, it is understood that the formation of steatosis is not latent even under conditions of a moderate shift in metabolic parameters.

Cluster 4 from cluster 2 is distinguished by pronounced hypertension, low insulin values, less compensation for carbohydrate metabolism, but also less severity of liver steatosis. In addition, alcohol abuse is the highest in this group, and the lowest adherence to dietary recommendations. The fact that dyslipidemia is isolated is also obvious, which is confirmed by the data of a large population study under the auspices of NHANES, which included more than 23 thousand Americans, in patients with hepatic pathology with high levels of transaminases lipid profiles with low LDL and high HDL can be recorded, which may be caused by a defect in the synthesis of lipoproteins or a violation of the synthetic function of the liver and a marker of latent hepatopathies [14].

Conclusions. 1. Clustering of patient examination results demonstrates a reliable distribution of groups according to the severity of liver steatosis.

2. In case of non-compliance with dietary recommendations and the use of alcohol even within acceptable limits, the progression of liver steatosis occurs even against the background of minimal metabolic disturbances

3. The presence of genetically determined hyperlipidemia in combination with insulin resistance is an unfavorable background in the implementation of the clinical manifestations of metabolic disorders.

REFERENCES

1. Estes, C.; Razavi, H.; Loomba, R.; Younossi, Z.; Sanyal, A.J. Modeling the epidemic of nonalcoholic fatty liver disease demonstrates an exponential increase in burden of disease. Hepatology 2018, 67, 123-133.

2. Younossi, Z.M.; Koenig, A.B.; Abdelatif, D.; Fazel, Y.; Henry, L.; Wymer, M. Global epidemiology of nonalcoholic fatty liver disease-Meta-analytic assessment of prevalence, incidence, and outcomes. Hepatology 2016, 64, 73-84.

3. Younossi, Z.; Anstee, Q.M.; Marietti, M.; Hardy, T.; Henry, L.; Eslam, M.; George, J.; Bugianesi, E. Global burden of NAFLD and NASH: Trends, predictions, risk factors and prevention. Nat. Rev. Gastroenterol. Hepatol. 2018, 15, 11-20.

4. Zelber-Sagi S, Nitzan-Kaluski D, Halpern Z, Oren R. Prevalence of primary non-alcoholic fatty liver disease in a population-based study and its association with biochemical and anthropometric measures. Liver Int 2006; 26: 856-863.

5. Chitturi S, Abeygunasekera S, Farrell GC, Holmes-Walker J, Hui JM, Fung C, Karim R, Lin R, Samarasinghe D, Liddle C, Weltman M, George J. NASH and insulin resistance: Insulin hypersecretion and specific association with the insulin resistance syndrome. Hepatology 2002; 35: 373-379.

6. Marchesini G, Marzocchi R, Agostini F, Bugianesi E. Nonalcoholic fatty liver disease and the metabolic syndrome. Curr Opin Lipidol 2005; 16: 421-427.

7. Estes, C.; Razavi, H.; Loomba, R.; Younossi, Z.; Sanyal, A.J. Modeling the epidemic of nonalcoholic fatty liver disease demonstrates an exponential increase in burden of disease. Hepatology 2018, 67, 123-133.

8. Kumari RSP, Vipula VA, Reddy BS, Nagadeepa W, Reddy BLN. Predictors of non-alcoholic fatty liver disease and non-alcoholic steatohepatitis in patients with type-2 diabetes mellitus. Int J Med Sci Public Health 2017;6:372-376.

9. Schilcher I, Ledinski G, Radulovic S, et al. Endothelial lipase increases antioxidative capacity of high-density lipoprotein. Biochim Biophys Acta Mol Cell Biol Lipids. 2019;1864(10):1363-1374. doi:10.1016/j.bbalip.2019.06.011

10. Junji Kobayashi. Which is the Best Predictor for the Development of Atherosclerosis Among Circulating Lipoprotein Lipase, Hepatic Lipase, and Endothelial Lipase? Journal of Atherosclerosis and Thrombosis, 2019, Volume 26, Issue 9: 758-759.

11. Kotronen A, Peltonen M, Hakkarainen A et al. Prediction of non-alcoholic fatty liver disease and liver fat using metabolic and genetic factors. Gastroenterology, vol. 137, no. 3: 865-872.

12. Sterling, R. K., Lissen, E. , Clumeck, N. , Sola, R. , Correa, M. C., Montaner, J. , S. Sulkowski, M., Torriani, F. J., Dieterich, D. T., Thomas, D. L., Messinger, D. and Nelson, M. Development of a simple noninvasive index to predict significant fibrosis in patients with HIV/HCV coinfection. Hepatology, 2006, 43: 1317-1325. doi:10.1002/hep.21178

13. Department of Health. UK Chief Medical Officers' Alcohol Guidelines Review: summary of the proposed new guidelines. London: Department of Health. 2016.

14. Jiang ZG, Mukamal K, Tapper E, Robson SC, Tsugawa Y Low LDL-C and High HDL-C Levels Are Associated with Elevated Serum Transaminases amongst Adults in the United States: A Cross-sectional Study. PLoS ONE, 2014,9(1):e85366. https://doi.org/10.1371/journal.pone.0085366