© 2013, Hepatobiliary Pancreat Dis Int. All rights reserved.

doi: 10.1016/S1499-3872(13)60091-0

Contributors: ZP proposed the study. XS and ZP performed research and wrote the first draft. LZW and TYF collected and analyzed the data. All authors contributed to the design and interpretation of the study and to further drafts. ZP is the guarantor.

Funding: None.

Ethical approval: Not needed.

Competing interest: No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article.

 

 

(Hepatobiliary Pancreat Dis Int 2013;12:575-583)

 

KEY WORDS: hepatitis B virus; hepatitis C virus; HBV status; pancreatic cancer; meta-analysis

Pancreatic cancer (PC) is one of the most lethal gastrointestinal cancers. The mortality to incidence ratio is 98%.^[1] Typically, due to the lack of early symptoms and accurate biomarkers, PC is usually diagnosed at an advanced stage when surgical resection is no longer possible. Unfortunately, the cause of PC remains largely unknown. Although many factors such as demographical factors, genetic factors, diabetes, consumption of alcohol and cigarette,^[2-4] have been reported to be involved in PC, we are far from having a comprehensive understanding of the risk factors. It is important to note that the identification of individuals who are at a high risk for PC may make the early diagnosis and treatment possible.

 

Approximately 350 million people worldwide are chronically infected by hepatitis B virus (HBV)^[5] and 180 million by hepatitis C virus (HCV).^[6] China is one of the endemic areas. HBV infection accounts for about 60% of the total liver cancer in developing countries and 23% in developed countries. The corresponding percentages for HCV infection are 33% in developing countries and 20% in developed countries.^[1] However, HBV and HCV are not only hepatotropic pathogens, they are also able to infect extrahepatic organs.^[7-9] To date, these viruses have been isolated from extrahepatic sites, such as the pancreas, kidney, skin, breast and blood vessel walls.^{[7, 10]} In addition, it has not previously considered that hepatitis viruses contribute to carcinogenesis of some organs apart from the liver. To date, it is clear that these pathogens may have an etiological role in some extrahepatic cancers, such as breast cancer,^[11] intrahepatic cholangiocarcinoma,^[12] and non-Hodgkin's lymphoma.^[13] Noteworthy, the liver and pancreas anatomically share common blood vessels and ducts, making the pancreas another potential target of hepatitis viruses.^[14]

 

In the last 5 years, studies^[15-23] were conducted to find the association between chronic hepatitis virus infection and the risk of PC. One of these studies showed inconsistent results because of various factors including study design, regional and racial difference, and sample size.^[15] Katakural et al^[8] in Japan found the association between chronic hepatitis virus infection and PC; however, Gonzalez et al^[24] in Korea did not find this association. The present meta-analysis was to answer the following question: Is chronic hepatitis virus infection (HBV and HCV) associated with PC? If yes, which HBV status is associated with PC?

 

 

The following databases were searched: Cochrane Library, PubMed, Ovid, Embase, Science Citation Index Expanded, and two Chinese databases (China National Knowledge Infrastructure and Wanfang database) (from January 2000 to July 2012). The following key words were used: pancreatic cancer or carcinoma or tumor and hepatitis B virus or hepatitis C virus or chronic hepatitis or HBV or HCV. Potentially relevant studies obtained via electronic search were extracted and their abstracts and full texts were considered for further evaluation. In case of duplicate reports, only the most recent version was considered in our analysis. Abstracts without full text or unpublished reports were excluded. When necessary, authors of included studies were contacted to obtain clarifications.

 

The titles and abstracts of all citations identified by the literature search were reviewed. Articles published in any language were included, if they fulfilled the following criteria: (1) studies were designed as cohort or case-control researches; (2) the diagnosis of PC was made by histology or imaging techniques; (3) HBV or HCV status was evaluated by viral markers, including hepatitis B surface antigen (HBsAg), anti-HBsAg antibody (HBsAb), anti-HBcAg antibody (HBcAb), hepatitis B e antigen (HBeAg), anti-HBeAg antibody (HBeAb), and HCV antibody (anti-HCV); (4) studies aimed to assess the possible relationship between chronic hepatitis and PC.

 

According to our predefined criteria, we excluded the followings from our analysis: (1) editorial, letters, news, reviews, expert opinions, case report and study without original data; (2) subset of published articles by the same author; (3) studies lacking control groups; (4) studies with unclear data; (5) studies involving subjects with acute hepatitis or pancreatitis, hepatocellular carcinoma (HCC), liver metastases; (6) studies enrolling HBV positive patients with hepatitis C-, hepatitis delta-, human immunodeficiency virus (HIV)-coinfection or alcohol-related hepatic or pancreatic diseases.

 

The quality of all selected papers was assessed independently by two authors (LZW and TYF). Discrepancies were resolved by discussion with a third investigator (ZP), or consensus of the whole team when necessary. The quality of studies was assessed using the scores proposed by modified Newcastle-Ottawa Quality Assessment Scale (NOS).^[25] The detail of modified NOS grading standard, including case-control or cohort study, was listed as follows: (i) selection, total score: 4; (ii) comparability, total score: 2; and (iii) exposure (case-control)/outcome (cohort), total score: 3. A high score out of a total of nine points indicates high methodological quality. A meta-regression analysis was performed to assess the possibility of confounding variables such as smoking status, alcohol use, and history of family cancer and diabetes.

 

In addition, HBV status was evaluated by HBeAg, HBcAb, HBsAb, HBeAb, and other five different patterns of HBV status. In patients with an increase of transaminases, the presence of HBsAg and anti-HCV as well as the detectable levels of HBV DNA were used as makers of chronic hepatitis virus infection. PC was diagnosed by histology, symptoms and signs and more than two types of imaging tools.

 

Two investigators (LZW and ZLM) extracted the data, utilizing a pre-designed data extraction form which met the inclusion criteria above. According to the Meta-analysis Of Observational Studies in Epidemiology (MOOSE),^[26] the outcomes were summarized from the ten studies. There was more than 98% agreement for data extraction between both investigators. The studies were classified according to their design. The following data, though some studies did not include all of them, were extracted from identified trials: (1) study characteristics including (a) first author's name, publication date, country of origin, and (b) the number of patients involved in retrospective or prospective studies and details of research design; and (2) patient demographics including (a) mean age, gender, matching variables such as diabetes, alcohol drinking, smoking, family cancer history, and (b) the prevalence of HBV or HCV seropositivity in case-control or cohort studies, and odds ratio (OR) or risk ratio (RR).

 

Freeware program review manager (version 5.0)^[27] and meta-disc^{[28, 29]} were used for data manipulation and statistical analysis. Measures of interest were OR for case-control studies, RR for cohort studies and the 95% confidence interval (CI). Case-control studies were considered with cohort studies evaluating virus infection and risk of PC. All statistical tests were two-sided. Fixed-effect model was used for meta-analysis unless there was significant heterogeneity. In this case, results were confirmed using a random-effects statistical model. The heterogeneity among trials evaluated by the Chi-square-based Q testing (Statistical significance cut-off for the test of heterogeneity was set at <0.10) and I² statistics (I²>50%). The random-effect model adjusted for variability of results among the trials provided a more conservative estimate of an effect using wider CI.

 

Publication bias was examined by a funnel plot of log(OR) against OR. If there was evidence of publication bias, the funnel plot would be asymmetric. Furthermore, we also conducted a sensitivity analysis, by excluding any single study each time. Meta-regression analysis was performed to assess the source of the heterogeneity and ruled out the effects of confounding factors.

 

 

The predefined search strategy yielded 214 potentially relevant papers from the published literature. Of these, we excluded 204 articles and the remaining 10 articles fulfilled the criteria for this systematic review. Selection of studies and reasons for exclusion are presented in Fig. 1. Eight of the 10 articles were case-control studies and two, cohort studies; 7 articles were from China including 1 from Taiwan province, 2 from Korea, and 1 from the United States. The main characteristics of these studies are summarized in Tables 1 and 2. All case studies were published in fulltext, and three of them were published in Chinese, six in English and one in Korean. Control subjects originated from hospital-based or general population-based design. Two cohort studies, one was from Taiwan province of China, another from Korea. Gonzalez's study,^[24] which described HBV status only in 32% of enrolled patients, was not included in our meta-analysis.

 

The quality assessment for all of studies according to the modified NOS is presented in Table 3. The assessment of different items, considered in NOS, produced the following mean values: comparability 2.7/4, selection 1.2/2, and outcome 2.1/3. Above all, case-control studies included 6951 subjects with PC, among these, 398 were HBsAg positive and 54 HCV positive. The available cohort studies included 3930 HBsAg and 146 394 HCV positive patients, of whom, 12 and 140 patients had PC, respectively. The most commonly reported risk factors for PC included: diabetes, alcohol consumption, smoking and family cancer history.

 

In our meta-analysis, we assessed the association between HBsAg positivity and the risk of PC development. Data were available in 8 case-control and 1 cohort studies, and we found that the presence of HBsAg correlated with a significantly increased risk of this malignancy (OR=1.28, 95% CI: 1.11-1.48, P=0.0003). No significant heterogeneity in this comparison was observed (P=0.15, I²=33%) (Fig. 2).

 

Then we evaluated the relationship between HCV and PC. Six case-control and 1 cohort studies showed that anti-HCV seropositivity was associated with a significantly increased PC (OR=1.21, 95% CI: 1.02-1.44, P=0.02). No significant heterogeneity in this comparison was detected (P=0.30, I²=17%) (Fig. 2).

 

We assessed whether other HBV antigen/antibody patterns are associated with PC, in addition to HBsAg. We analyzed 6 studies and found that serum HBsAb and HBeAb correlated with a lower PC risk in comparison with the absence of serum HBsAb/HBeAb (OR=0.40, 95% CI: 0.20-0.79, P=0.008 and OR=0.62, 95% CI: 0.39-0.99, P=0.04, respectively). On the other hand, the presence of serum HBcAb and HBeAg was not associated with a statistically significant increase of PC (OR=1.10, 95% CI: 0.77-1.57, P=0.59 and OR=1.75, 95% CI: 0.77-4.01, P=0.18, respectively). However, a significant heterogeneity in this comparison was detected (Fig. 3).

 

The presence of HBsAb and HBeAb was protective and they were negatively correlated with PC. HBsAg positivity was found to increase the risk of PC. According to clinical practice and available studies, HBsAg^– patients with HBsAb and/or HBcAb positivity were grouped together and divided into five subgroups, including HBsAg^–/HBcAb⁺, HBsAb⁺/HBcAb⁺, HBsAb^–/HBcAb⁺, HBsAg^–/HBcAb⁺/HBsAb^–, and HBsAg^–/HBcAb⁺/HBsAb⁺. Interestingly, as shown in Fig. 4, results of our meta-analysis indicated that these viral antigen/antibody patterns are not associated with a higher risk of PC. However, all subgroups presented a significant heterogeneity in our analysis (Fig. 4).

 

The influence of the publication(s) with statistically significant heterogeneity on the overall meta-analysis estimate was investigated. We performed a meta-regression analysis to assess the source of the heterogeneity and ruled out the effects of confounding factors (Fig. 5). The effect of each study on the overall meta-analysis estimate was assessed by excluding one study at a time, but this approach produced no alteration of final decision, demonstrating that our results were reliable.

 

The funnel plot did not suggest the presence of publication bias in our meta-analysis (Fig. 6).

 

 

There are several documentations about the frequency and characteristics of pancreatic disorder and chronic viral hepatitis. More than 3 decades ago, Japanese researchers^[30] found HBV in pancreatic endocrine and exocrine tissue by immunohistochemistry and electronic microscopy. Later on, studies^{[8, 31, 32]} suggested that both HBV and HCV can replicate in the pancreas and cause inflammation, persistent damage and subsequent possible malignant transformation. This fascinated a lot of researchers up to now. In addition, compared with virus-associated HCC, viral DNA integration into the genome of pancreatic host cells may cause a deregulation of key regulators of cell control, and induce tumor progression.^[33] To date, many authors have performed a lot of investigations on the association between chronic viral hepatitis and PC risk. However, because of the different study design and population enrolled, no definitive conclusions have been made up to now.

 

Being different from most of the previous studies, our meta-analysis has certain innovation. This analysis absorbed the idea of a systematic review^[34] on the possible role of HBV and HCV as risk factors for PC and was influenced by three meta-analyses on this topic.^[35-37] Wang et al^[35] and Luo et al^[36] only analyzed the association between active or previous HBV infection and PC. They reported that HBV might enhance the probability of this carcinoma. On the other hand, Fiorino et al assessed, for the first time, PC risk in patients with HBV/HCV chronic infection.^[34] He reported a higher risk of PC in HBsAg positive carriers, whereas the relationship between PC risk and anti-HCV reached a borderline but was not of significant value, because only 3 studies were available at that time. In our meta-analysis we found a marked association between HCV infection and PC and an increased probability of PC in patients with chronic HBV infection. Furthermore, we assessed whether the presence of other serum HBV markers, including HBeAg, HBsAb, HBcAb and HBeAb, or different viral antigen/antibody patterns (HBsAg^–/HBcAb⁺, HBcAb⁺/HBsAb⁺, HBcAb⁺/HBsAb^–) correlated with an increased or decreased risk of this malignancy. In our analysis, the presence of HBsAb or HBeAb was associated with a decreased risk of PC, but the heterogeneity in these comparisons was high. Several studies^{[14, 33, 36]} have assessed different anti-HBV antibody patterns such as HBcAb⁺/HBsAb^–, HBcAb⁺/HBsAg⁺ and HBcAb⁺/HBsAb⁺. However, the overall data are inconclusive. In our analysis we analyzed two HBV antigen/antibody combinations, they were HBsAg^–/HBcAb⁺/HBsAb^– and HBsAg^–/HBcAb⁺/HBsAb⁺. These profiles have been considered as signs of a complete recovery from a past exposure to HBV infection. We concluded that these viral antigen/antibody patterns is not associated with a higher risk of PC.^{[14, 16, 33]}

 

There are limitations in our study. First, almost all studies were performed in Asian populations and, in particular, in China, where viral liver infections are more prevalent. Therefore, the validity of our meta-analysis is limited to Asian people. Furthermore, our report presents the possibility of a type II error, caused by multiple end-points. Second, our article includes studies with heterogeneous study design, enrolling populations from different geographic areas. In addition, some studies were performed with the primary aim to investigate overall risk factors of pancreatic cancer, natural history of chronic hepatitis B, risk predictors of disease progression and HCC development. Because of the possibility of confounding variables such as smoking status, alcohol use and history of diabetes and family cancer, we performed a meta-regression analysis to assess these factors. We found that the risk of PC is independent of these variables. Third, most of the patients were from epidemic regions where the prevalence of HBV or HCV infection is higher than other geographic areas, therefore markers of either past or active HBV and/or HCV infection are frequently detected also in subjects with PC, so the relationship between HBV and/or HCV infection and PC could be a casual association. In order to minimize the risk of bias, we developed a detailed protocol before starting our study. According to the MOOSE system, we performed carefully an analysis of selected studies by using the modified NOS, and credible evidences suggested the existence of an association between chronic hepatitis virus infection and PC.

 

Our research has important clinical consequence. To date, several reasons have prevented large-scale studies, which could improve our knowledge of causes and mechanisms involved in pancreatic carcinogenesis. Our study may contribute to identify subjects with a high risk of PC. Therefore a general approach based on screening and anti-HBV vaccination programs might be a part of prevention strategies against PC.

 

In summary, our meta-analysis, including 8 case-control studies and 2 cohort studies, confirms that HBV or HCV infection is associated with an increased risk of PC in HBsAg and anti-HCV positive patients. Furthermore, our study suggests that the presence of HBsAb or HBeAb may be a protective factor of PC. It is still uncertain whether serological pattern of past exposure to HBV with or without natural immunity is associated with an increased probability of this malignancy. Further well-designed multicenter trials of long duration with an adequate number of patients are needed to confirm our results.

 

 

1 Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. Global cancer statistics. CA Cancer J Clin 2011;61:69-90. PMID: 21296855

2 Krechler T, Novotny J, Zeman M, Krska Z, Svestka T, Sváb J, et al. Pancreatic carcinoma and diabetes mellitus. Cas Lek Cesk 2004;143:97-100. PMID: 15077571

3 Li D, Morris JS, Liu J, Hassan MM, Day RS, Bondy ML, et al. Body mass index and risk, age of onset, and survival in patients with pancreatic cancer. JAMA 2009;301:2553-2562. PMID: 19549972

4 Liao KF, Lai SW, Li CI, Chen WC. Diabetes mellitus correlates with increased risk of pancreatic cancer: a population-based cohort study in Taiwan. J Gastroenterol Hepatol 2012;27:709-713. PMID: 21929650

5 Boglione L, D'Avolio A, Cariti G, Gregori G, Burdino E, Baietto L, et al. Kinetics and prediction of HBsAg loss during therapy with analogues in patients affected by chronic hepatitis B HBeAg negative and genotype D. Liver Int 2013;33:580-585. PMID: 23311449

6 Kim DY, Kim IH, Jeong SH, Cho YK, Lee JH, Jin YJ, et al. A nationwide seroepidemiology of hepatitis C virus infection in South Korea. Liver Int 2013;33:586-594. PMID: 23356674

7 Dejean A, Lugassy C, Zafrani S, Tiollais P, Brechot C. Detection of hepatitis B virus DNA in pancreas, kidney and skin of two human carriers of the virus. J Gen Virol 1984;65: 651-655. PMID: 6699625

8 Katakura Y, Yotsuyanagi H, Hashizume K, Okuse C, Okuse N, Nishikawa K, et al. Pancreatic involvement in chronic viral hepatitis. World J Gastroenterol 2005;11:3508-3513. PMID: 15962364

9 Yan FM, Chen AS, Hao F, Zhao XP, Gu CH, Zhao LB, et al. Hepatitis C virus may infect extrahepatic tissues in patients with hepatitis C. World J Gastroenterol 2000;6:805-811. PMID: 11819700

10 Trepo CG, Zucherman AJ, Bird RC, Prince AM. The role of circulating hepatitis B antigen/antibody immune complexes in the pathogenesis of vascular and hepatic manifestations in polyarteritis nodosa. J Clin Pathol 1974;27:863-868. PMID: 4155412

11 Su FH, Chang SN, Chen PC, Sung FC, Su CT, Yeh CC. Association between chronic viral hepatitis infection and breast cancer risk: a nationwide population-based case-control study. BMC Cancer 2011;11:495. PMID: 22115285

12 Zhou YM, Zhao YF, Li B, Huang JY, Wu LP, Xu DH, et al. Hepatitis viruses infection and risk of intrahepatic cholangiocarcinoma: evidence from a meta-analysis. BMC Cancer 2012;12:289. PMID: 22799744

13 Ulcickas Yood M, Quesenberry CP Jr, Guo D, Caldwell C, Wells K, Shan J, et al. Incidence of non-Hodgkin's lymphoma among individuals with chronic hepatitis B virus infection Hepatology 2007;46:107-112. PMID: 17526021

14 Wang DS, Chen DL, Ren C, Wang ZQ, Qiu MZ, Luo HY, et al. ABO blood group, hepatitis B viral infection and risk of pancreatic cancer. Int J Cancer 2012;131:461-468. PMID: 21858814

15 Lai SW, Liao KF. Hepatitis B and pancreatic cancer. Eur J Cancer Prev 2013;22:620. PMID: 24061269

16 Ben Q, Li Z, Liu C, Cai Q, Yuan Y, Wang K, et al. Hepatitis B virus status and risk of pancreatic ductal adenocarcinoma: a case-control study from China. Pancreas 2012;41:435-440. PMID: 22422136

17 Zhu F, Li HR, Du GN, Chen JH, Cai SR. Chronic hepatitis B virus infection and pancreatic cancer: a case-control study in southern China. Asian Pac J Cancer Prev 2011;12:1405-1408. PMID: 22126472

18 Xu P, Huang Q, Liu CH, Xie F, Shao F, Zhu CL, et al. Risk factors for pancreatic cancer: a case-control study. Tumor (Chinese) 2011;31:653-657.

19 Li HR, Lin TY. Hepatitis B virus infection in pancreatic cancer and lymphoma patients, comparison of entecavir and lamivudine in preventing hepatitis B reactivation in lymphoma patients during chemotherapy. Master's dissertation of Sun Yat-Sen University in Guangzhou (Chinese) 2010.

20 Hong SG, Kim JH, Lee YS, Yoon E, Lee HJ, Hwang JK, et al. The relationship between hepatitis B virus infection and the incidence of pancreatic cancer: a retrospective case-control study. Korean J Hepatol 2010;16:49-56.

21 Ma W, Xie DR, Cao WP, Yang Q, Jiang ZM, Chen DL, et al. Association between hepatitis B virus infection and pancreatic cancer. Chin J Clin Oncol (Chinese) 2009;36:1388-1390.

22 Iloeje UH, Yang HI, Jen CL, Su J, Wang LY, You SL, et al. Risk of pancreatic cancer in chronic hepatitis B virus infection: data from the REVEAL-HBV cohort study. Liver Int 2010;30: 423-429. PMID: 19840258

23 El-Serag HB, Engels EA, Landgren O, Chiao E, Henderson L, Amaratunge HC, et al. Risk of hepatobiliary and pancreatic cancers after hepatitis C virus infection: A population-based study of U.S. veterans. Hepatology 2009;49:116-123. PMID: 19085911

24 Berrington de Gonzalez A, Yun JE, Lee SY, Klein AP, Jee SH. Pancreatic cancer and factors associated with the insulin resistance syndrome in the Korean cancer prevention study. Cancer Epidemiol Biomarkers Prev 2008;17:359-364. PMID: 18268120

25 Wells GA, Shea B, O'Connell D, Peterson J, Welch V, Losos M, et al. The New castle-Ottawa Scale (NOS) for assessing the quality of nonrandomized studies in meta-analyses [serial online]. Accessed January 1, 2008. Available at http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp

26 Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D, et al. Meta-analysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies in Epidemiology (MOOSE) group. JAMA 2000;283:2008-2012. PMID: 10789670

27 Higgins JPT, Green S. Cochrane Handbook for Systematic Reviews of Interventions [serial online]. Accessed March 1, 2011. Available from: http://www.mrc-bsu.cam.ac.uk/cochrane/handbook/

28 Zamora J, Abraira V, Muriel A, Khan K, Coomarasamy A. Meta-DiSc: a software for meta-analysis of test accuracy data. BMC Med Res Methodol 2006;6:31. PMID: 16836745

29 U. de Bioestadística-Software: Meta-Disc. http://www.hrc.es/ investigacion/ metadisc_en.htm.

30 Yoshimura M, Sakurai I, Shimoda T, Abe K, Okano T, Shikata T. Detection of HBsAg in the pancreas. Acta Pathol Jpn 1981;31:711-717. PMID: 7025575

31 Taranto D, Carrato A, Romano M, Maio G, Izzo CM, Del Vecchio Blanco C. Mild pancreatic damage in acute viral hepatitis. Digestion 1989;42:93-97. PMID: 2475380

32 Raza S, Chaudhry NA, Brown JD, Aghaie S, Rezai D, Khan A, et al. To Study the Clinical, Biochemical and Radiological Features of Acute Pancreatitis in HIV and AIDS. J Clin Med Res 2013;5:12-17 PMID: 23390470

33 Hassan MM, Li D, El-Deeb AS, Wolff RA, Bondy ML, Davila M, et al. Association between hepatitis B virus and pancreatic cancer. J Clin Oncol 2008;26:4557-4562. PMID: 18824707

34 Fiorino S, Lorenzini S, Masetti M, Deleonardi G, Grondona AG, Silvestri T, et al. Hepatitis B and C virus infections as possible risk factor for pancreatic adenocarcinoma. Med Hypotheses 2012;79:678-697. PMID: 22959312

35 Wang Y, Yang S, Song F, Cao S, Yin X, Xie J, et al. Hepatitis B virus status and the risk of pancreatic cancer: a meta-analysis. Eur J Cancer Prev 2013;22:328-334. PMID: 23165286

36 Luo G, Hao NB, Hu CJ, Yong X, Lü MH, Cheng BJ, et al. HBV infection increases the risk of pancreatic cancer: a meta-analysis. Cancer Causes Control 2013;24:529-537. PMID: 23306552

37 Fiorino S, Chili E, Bacchi-Reggiani L, Masetti M, Deleonardi G, Grondona AG, et al. Association between hepatitis B or hepatitis C virus infection and risk of pancreatic adenocarcinoma development: a systematic review and meta-analysis. Pancreatology 2013;13:147-160. PMID: 23561973