Prevalence and Predictors of Esophageal Varices in Patients With Primary Biliary Cirrhosis
2007; Elsevier BV; Volume: 5; Issue: 7 Linguagem: Inglês
10.1016/j.cgh.2007.02.031
ISSN1542-7714
AutoresCynthia Levy, Claudia O. Zein, Justin M. Gomez, Consuelo Soldevila‐Pico, Roberto J. Firpi, Giuseppe Morelli, David R. Nelson,
Tópico(s)Liver Diseases and Immunity
ResumoBackground & Aims: Esophageal varices and bleeding are among the most feared complications of primary biliary cirrhosis (PBC). We aimed to determine the prevalence of esophageal varices in patients with PBC, to evaluate noninvasive markers of esophageal varices in this population, and to validate the results in an independent set of patients. Methods: Data were collected on all patients with PBC seen for the first time at the University of Florida (study group) and at Case Western Reserve University hospitals (cross-validation group) during 7 consecutive years. Logistic regression analysis was used to identify independent predictors of esophageal varices. The best cut-off values were calculated based on receiver operating characteristic curves. The diagnostic accuracy of the independent predictors of esophageal varices identified in the study group were retested in the cross-validation group. Results: Of 210 patients with PBC seen at the University of Florida, 113 had an endoscopy and 49.6% (56 of 113) were found to have esophageal varices. After excluding 22 patients with a history of variceal bleeding, data on 91 patients were analyzed. Thirty-four patients had esophageal varices (37%). Multivariate analysis revealed that a platelet count of less than 140,000 (odds ratio, 7.6; 95% confidence interval, 1.6–37) and a Mayo risk score of 4.5 or greater (odds ratio, 10.6; 95% confidence interval, 1.8–62) were independent predictors of esophageal varices. The diagnostic accuracy of these predictors was confirmed in an independent set of patients. Conclusions: Among patients with PBC, a platelet count of less than 140,000 and/or a Mayo risk score of 4.5 or greater appears to identify those patients more likely to benefit from a screening endoscopy. Background & Aims: Esophageal varices and bleeding are among the most feared complications of primary biliary cirrhosis (PBC). We aimed to determine the prevalence of esophageal varices in patients with PBC, to evaluate noninvasive markers of esophageal varices in this population, and to validate the results in an independent set of patients. Methods: Data were collected on all patients with PBC seen for the first time at the University of Florida (study group) and at Case Western Reserve University hospitals (cross-validation group) during 7 consecutive years. Logistic regression analysis was used to identify independent predictors of esophageal varices. The best cut-off values were calculated based on receiver operating characteristic curves. The diagnostic accuracy of the independent predictors of esophageal varices identified in the study group were retested in the cross-validation group. Results: Of 210 patients with PBC seen at the University of Florida, 113 had an endoscopy and 49.6% (56 of 113) were found to have esophageal varices. After excluding 22 patients with a history of variceal bleeding, data on 91 patients were analyzed. Thirty-four patients had esophageal varices (37%). Multivariate analysis revealed that a platelet count of less than 140,000 (odds ratio, 7.6; 95% confidence interval, 1.6–37) and a Mayo risk score of 4.5 or greater (odds ratio, 10.6; 95% confidence interval, 1.8–62) were independent predictors of esophageal varices. The diagnostic accuracy of these predictors was confirmed in an independent set of patients. Conclusions: Among patients with PBC, a platelet count of less than 140,000 and/or a Mayo risk score of 4.5 or greater appears to identify those patients more likely to benefit from a screening endoscopy. Variceal bleeding is among the most serious complications of portal hypertension. Esophageal varices are present in roughly 50% of all cirrhotic patients, with a 2%–15% annual incidence of bleeding.1D'Amico G. Luca A. Natural history Clinical-haemodynamic correlations. Prediction of the risk of bleeding.Baillieres Clin Gastroenterol. 1997; 11: 243-256Abstract Full Text PDF PubMed Scopus (200) Google Scholar Despite recent advances in medical management, the probability of death after the first episode of bleeding remains 15%–21%.2McCormick P.A. O'Keefe C. Improving prognosis following a first variceal haemorrhage over four decades.Gut. 2001; 49: 682-685Crossref PubMed Scopus (179) Google Scholar, 3El-Serag H.B. Everhart J.E. Improved survival after variceal hemorrhage over an 11-year period in the Department of Veterans Affairs.Am J Gastroenterol. 2000; 95: 3566-3573Crossref PubMed Google Scholar, 4D'Amico G. De Franchis R. Upper digestive bleeding in cirrhosis Post-therapeutic outcome and prognostic indicators.Hepatology. 2003; 38: 599-612Crossref PubMed Scopus (681) Google Scholar Such high mortality combined with the efficacy of primary prophylaxis, which leads to a 40% reduction in the risk of death, reinforce the need for a systematic approach to detect varices in at-risk patients.5D'Amico G. Pagliaro L. Bosch J. Pharmacological treatment of portal hypertension: an evidence-based approach.Semin Liver Dis. 1999; 19: 475-505Crossref PubMed Scopus (629) Google Scholar, 6Grace N.D. Groszmann R.J. Garcia-Tsao G. et al.Portal hypertension and variceal bleeding: an AASLD single topic symposium.Hepatology. 1998; 28: 868-880Crossref PubMed Scopus (330) Google Scholar Practice guidelines endorsed by the American Association for the Study of Liver Diseases in 1998 suggest that all newly diagnosed cirrhotic patients should be screened for esophageal varices; those with large varices should be treated with β-blockers, and those with small varices should be rescreened in approximately 1 year.6Grace N.D. Groszmann R.J. Garcia-Tsao G. et al.Portal hypertension and variceal bleeding: an AASLD single topic symposium.Hepatology. 1998; 28: 868-880Crossref PubMed Scopus (330) Google Scholar These recommendations, however, may not apply to patients with primary biliary cirrhosis (PBC), who may develop portal hypertension without frank cirrhosis.PBC is a chronic cholestatic liver disease characterized by progressive destruction of interlobular bile ducts and chronic cholestasis, eventually leading to cirrhosis and its complications.7Talwalkar J.A. Lindor K.D. Primary biliary cirrhosis.Lancet. 2003; 362: 53-61Abstract Full Text Full Text PDF PubMed Scopus (289) Google Scholar Approximately one third of patients with PBC develop esophageal varices during follow-up evaluation, and half of these will have a documented episode of variceal bleeding.8Kew M.C. Varma R.R. Dos Santos H.A. et al.Portal hypertension in primary biliary cirrhosis.Gut. 1971; 12: 830-834Crossref PubMed Scopus (84) Google Scholar, 9Gores G.J. Wiesner R.H. Dickson E.R. et al.Prospective evaluation of esophageal varices in primary biliary cirrhosis: development, natural history, and influence on survival.Gastroenterology. 1989; 96: 1552-1559Abstract PubMed Google Scholar More recent studies, however, have indicated a lower prevalence of varices (8%–19%), perhaps owing to the long-term use of ursodeoxycholic acid (UDCA).10Angulo P. Lindor K. Therneau T. et al.Utilization of the Mayo risk score in patients with primary biliary cirrhosis receiving ursodeoxycholic acid.Liver. 1999; 19: 115-121Crossref PubMed Scopus (137) Google Scholar, 11Lindor K.D. Jorgensen R.A. Therneau T.M. et al.Ursodeoxycholic acid delays the onset of esophageal varices in primary biliary cirrhosis.Mayo Clin Proc. 1997; 72: 1137-1140Abstract Full Text Full Text PDF PubMed Scopus (121) Google Scholar, 12Jones D. Walter R. Prince M. et al.The prevalence of portal hypertension in primary biliary cirrhosis.Gut. 2002; 50: A113Google Scholar Despite a strong correlation between advanced stages of PBC and the presence of esophageal varices, portal hypertension can occur during early stages of the disease. In addition to sampling error, several explanations exist for this occurrence, including significant portal tract inflammation causing portal venous compression, perisinusoidal fibrosis, and nodular regenerative hyperplasia.13Sherlock S. Noncirrhotic extrahepatic and intrahepatic portal hypertension.Semin Liver Dis. 1982; 2: 202-210Crossref PubMed Scopus (16) Google Scholar, 14Nakanuma Y. Ohta G. Nodular hyperplasia of the liver in primary biliary cirrhosis of early histological stages.Am J Gastroenterol. 1987; 82: 8-10PubMed Google Scholar, 15Navasa M. Pares A. Bruguera M. et al.Portal hypertension in primary biliary cirrhosis Relationship with histological features.J Hepatol. 1987; 5: 292-298Abstract Full Text PDF PubMed Scopus (76) Google Scholar, 16Colina F. Pinedo F. Solis J.A. et al.Nodular regenerative hyperplasia of the liver in early histological stages of primary biliary cirrhosis.Gastroenterology. 1992; 102: 1319-1324Abstract Full Text PDF PubMed Google Scholar Thus, in this population, complications of portal hypertension can occur before the development of cirrhosis. Further complicating the matter, a liver biopsy examination is not always necessary to make a diagnosis of PBC,17Zein C. Angulo P. Lindor K. When is liver biopsy needed in the diagnosis of primary biliary cirrhosis?.Clin Gastroenterol Hepatol. 2003; 1: 89-95Abstract Full Text Full Text PDF PubMed Scopus (71) Google Scholar nor is it included in the most commonly used prognostic index for this disease, the Mayo risk score.18Dickson E.R. Grambsch P.M. Fleming T.R. et al.Prognosis in primary biliary cirrhosis: model for decision making.Hepatology. 1989; 10: 1-7Crossref PubMed Scopus (670) Google Scholar Therefore, many experts are moving away from performing diagnostic and follow-up liver biopsy procedures in PBC, which reduces the likelihood of diagnosing cirrhosis.Although several groups have attempted to define noninvasive predictors of esophageal varices among cirrhotic patients,19Cottone M. D'Amico G. Maringhini A. et al.Predictive value of ultrasonography in the screening of non-ascitic cirrhotic patients with large varices.J Ultrasound Med. 1986; 5: 189-192PubMed Google Scholar, 20Chalasani N. Imperiale T.F. Ismail A. et al.Predictors of large esophageal varices in patients with cirrhosis.Am J Gastroenterol. 1999; 94: 3285-3291Crossref PubMed Google Scholar, 21Pilette C. Oberti F. Aube C. et al.Non-invasive diagnosis of esophageal varices in chronic liver diseases.J Hepatol. 1999; 31: 867-873Abstract Full Text Full Text PDF PubMed Scopus (137) Google Scholar, 22Zaman A. Hapke R. Flora K. et al.Factors predicting the presence of esophageal or gastric varices in patients with advanced liver disease.Am J Gastroenterol. 1999; 94: 3292-3296Crossref PubMed Google Scholar, 23Madhotra R. Mulcahy H.E. Willner I. et al.Prediction of esophageal varices in patients with cirrhosis.J Clin Gastroenterol. 2002; 34: 81-85Crossref PubMed Scopus (167) Google Scholar, 24Schepis F. Camma C. Niceforo D. et al.Which patients with cirrhosis should undergo endoscopic screening for esophageal varices detection?.Hepatology. 2001; 33: 333-338Crossref PubMed Scopus (248) Google Scholar, 25Giannini E. Botta F. Borro P. et al.Platelet count/spleen diameter ratio: proposal and validation of a non-invasive parameter to predict the presence of oesophageal varices in patients with liver cirrhosis.Gut. 2003; 52: 1200-1205Crossref PubMed Scopus (338) Google Scholar, 26Thomopoulos K.C. Labropoulou-Karatza C. Mimidis K.P. et al.Non-invasive predictors of the presence of large oesophageal varices in patients with cirrhosis.Dig Liver Dis. 2003; 35: 473-478Abstract Full Text Full Text PDF PubMed Scopus (129) Google Scholar, 27Zein C.O. Lindor K.D. Angulo P. Prevalence and predictors of esophageal varices in patients with primary sclerosing cholangitis.Hepatology. 2004; 39: 204-210Crossref PubMed Scopus (97) Google Scholar a consistent and reproducible marker has not been identified for use in clinical practice.28D'Amico G. Morabito A. Noninvasive markers of esophageal varices: another round, not the last.Hepatology. 2004; 39: 30-34Crossref PubMed Scopus (68) Google Scholar In addition, patients with PBC were largely underrepresented in these studies. Angulo et al10Angulo P. Lindor K. Therneau T. et al.Utilization of the Mayo risk score in patients with primary biliary cirrhosis receiving ursodeoxycholic acid.Liver. 1999; 19: 115-121Crossref PubMed Scopus (137) Google Scholar evaluated data from 180 PBC patients who participated in a clinical trial, of whom 138 had screening esophagogastroduodenoscopy (EGD). The only noninvasive independent predictor of varices was the Mayo risk score. Ninety-three percent of patients who developed varices had a Mayo risk score of 4 or greater. More recently, Bressler et al examined a subgroup of 86 patients with PBC (n = 79) or primary sclerosing cholangitis (n = 7) who underwent EGD and found that a platelet count of less than 200,000/mm3, an albumin level of less than 4.0 g/dL, and a bilirubin level greater than 20 μmol/L (1.16 mg/dL) were independent predictors of esophageal varices.29Bressler B. Pinto R. El-Ashry D. et al.Which patients with primary biliary cirrhosis or primary sclerosing cholangitis should undergo endoscopic screening for oesophageal varices detection?.Gut. 2005; 54: 407-410Crossref PubMed Scopus (58) Google Scholar For those patients with all 3 variables, the probability of varices was 88%. Nevertheless, the predictive value of each variable was not discussed and the external validity of these tests was not assessed.The availability of noninvasive predictors of esophageal varices would allow for patient selection even in the absence of histology data, and would facilitate the decision-making process regarding the timing of a screening EGD in patients who had a liver biopsy performed several years prior, especially in view of the delayed progression of PBC in patients treated with UDCA. Also, strong evidence exists to suggest that portal hypertension may develop in precirrhotic patients with PBC. Given the fact that a liver biopsy may not be available or needed for every patient with PBC, and that universal primary prophylaxis with β-blockers is not an effective strategy,30Groszmann R.J. Garcia-Tsao G. Bosch J. et al.Beta-blockers to prevent gastroesophageal varices in patients with cirrhosis.N Engl J Med. 2005; 353: 2254-2261Crossref PubMed Scopus (712) Google Scholar noninvasive markers of esophageal varices still are desired. Thus, the primary aims of our study were as follows: (1) to determine the prevalence of esophageal varices in patients with PBC, (2) to evaluate noninvasive markers of esophageal varices in such a population, and (3) to validate our results in an independent set of patients with PBC.Patients and MethodsPatient PopulationA total of 210 patients with a diagnosis of PBC seen at the University of Florida (Gainesville, FL) for the first time between January 1, 1998, and December 31, 2004, were included in the present study. The year 1998 was selected because UDCA was approved by the Food and Drug Administration as standard therapy for PBC on December 1997, and previous studies have indicated that this drug may delay the development of esophageal varices.11Lindor K.D. Jorgensen R.A. Therneau T.M. et al.Ursodeoxycholic acid delays the onset of esophageal varices in primary biliary cirrhosis.Mayo Clin Proc. 1997; 72: 1137-1140Abstract Full Text Full Text PDF PubMed Scopus (121) Google Scholar Of these 210 patients, 113 underwent EGD: 91 for screening and 22 during evaluation for gastrointestinal bleeding. These 113 patients were included in our study population.An independent set of patients from Case Western Reserve University hospitals was used to validate our results. By using a medical index search engine, all patients seen for the first time at the University Hospitals of Cleveland and at Metro Health Medical Center in Cleveland, Ohio, between October 1998 and October 2005 who had a diagnosis of PBC were identified. Individual charts were reviewed and the diagnosis of PBC was confirmed in 92 patients. Thirty-six patients with PBC who had no history of esophageal varices bleeding and who had undergone EGD were included in the cross-validation set.For this study, a diagnosis of PBC required either histologic confirmation and the presence of chronic cholestasis with or without a positive antimitochondrial antibody (AMA), or a highly positive AMA and chronic cholestasis, even in the absence of a previous liver biopsy. The following information was abstracted for each patient: demographic data, date of diagnosis, liver biopsy results, date and results of EGD, laboratory tests within 3 months of the EGD, history of variceal hemorrhage, imaging studies (computed tomography, magnetic resonance, or ultrasound), and data from clinic visits. Only 1 EGD was entered per patient—either the latest performed or the first at which varices were seen. This study was approved by the University of Florida and Case Western Reserve University Institutional Review Boards.DefinitionsThe Mayo risk score was calculated for each patient at the time of his or her EGD. This score was defined as follows: 0.871 × loge (bilirubin in mg/dL) + (−) 2.53 × loge (albumin in g/dL) + 0.039 age in years + 2.83 loge (prothrombin time in seconds) + 0.859 edema.18Dickson E.R. Grambsch P.M. Fleming T.R. et al.Prognosis in primary biliary cirrhosis: model for decision making.Hepatology. 1989; 10: 1-7Crossref PubMed Scopus (670) Google Scholar The Model for End-stage Liver Disease score also was calculated at the time of the EGD. The model is as follows: 0.957 × loge (creatinine mg/dL) + 0.378 × loge (bilirubin mg/dL) + 1.120 × loge (international normalized ratio) + 0.643.31Kamath P.S. Wiesner R.H. Malinchoc M. et al.A model to predict survival in patients with end-stage liver disease.Hepatology. 2001; 33: 464-470Crossref PubMed Scopus (3667) Google ScholarStatistical AnalysisTo determine the prevalence of esophageal varices in PBC, all 113 patients were analyzed. Patients with a known history of variceal hemorrhage were excluded from univariate or multivariate analyses to predict the presence of varices. Descriptive statistics were used to characterize study patients. Associations between potential predictive variables and esophageal varices were measured using the χ2 or the Student t test. The Wilcoxon signed rank test was used for nonparametric variables. Multivariate logistic regression, stepwise forward, was performed to identify independent predictors of esophageal varices. Two-sided P values less than .05 were considered statistically significant. Variables in the multiple logistic regression analysis were used to create a probability prediction formula. Also, receiver operating characteristic curves and sensitivity and specificity plots were constructed for continuous variables found to be independent predictors of varices. Cut-off values providing the best sensitivity and specificity were identified. The predictive value of the variables identified in the original study group was assessed in the cross-validation group by contingency and logistic regression analysis.ResultsPrevalence of Esophageal VaricesOf 113 patients with PBC who underwent EGD, 56 (49.6%) had esophageal varices. Among 91 patients undergoing screening EGD, 34 (37%) were found to have varices. No statistically significant difference was noted regarding demographics, serum liver biochemistries, platelet count, prothrombin time, or AMA status between patients who underwent endoscopy and those who did not (data not shown). The clinical, biochemical, and histologic data of the 91 patients with PBC included in the analyses for predictors of varices are summarized in Table 1. Liver biopsy specimens were available for 80 of the 91 patients presenting for screening EGD. Interestingly, 7 of the 26 patients found to have esophageal varices had stages 1–2 disease according to liver biopsy specimens obtained an average of 5.7 years from the date of EGD. Notably, 2 patients had documented early histologic stage within 1–2 years of endoscopy.Table 1Characteristics of 127 Patients With PBC Without Previous Variceal BleedingCharacteristicsStudy group (n = 91)Cross-validation group (n = 36)P valueFemales84 (92%)29 (81%).04Age, y58 ± 1.157.5 ± 1.9NSAMA titer ≥ 1:40aAMA titer was not available for 5 patients in the study group and 3 patients in the cross-validation group.62/86 (72.1%)32/33 (97%).001Serum alkaline phosphatase level (normal range, 35–129 U/L)274 ± 25232 ± 57NSSerum bilirubin level (normal range, 0.0–1.0 mg/dL)0.9 ± 0.51.2 ± 0.5NSSerum aspartate aminotransferase level (normal range, 0–37 U/mL)61 ± 5.859 ± 14NSSerum alanine aminotransferase level (normal range, 0–41 U/mL)56 ± 7.058 ± 8.4NSSerum albumin level (normal range, 3.5–5 mg/dL)3.8 ± 0.13.8 ± 0.1NSSerum creatinine level (normal range, 0.8–1.2 mg/dL)0.8 ± 0.00.9 ± 0.1NSProthrombin time (normal range, 10.5–13.5 s)11.6 ± 0.211.3 ± 0.3NSPlatelet count (normal range, 150–450 thou/cu mm)155 ± 11193 ± 15NSMayo risk score4.6 ± 0.25.0 ± 0.4NSMELD score9 ± 0.6Esophageal varices None57/91 (62.6%)19/36 (52.8%)NS Small15/91 (16.5%)7/36 (19.4%)NS Large19/91 (20.8%)10/36 (27.8%)NSHistologic stagebLiver biopsy was available for 80 patients in the study group and 33 patients in the cross-validation group. 1–238/80 (47.5%)12/33 (36.4%)NS 3–442/80 (52.5%)21/33 (63.6%)NSNOTE. Values expressed are median ± SE or number (%).a AMA titer was not available for 5 patients in the study group and 3 patients in the cross-validation group.b Liver biopsy was available for 80 patients in the study group and 33 patients in the cross-validation group. Open table in a new tab Four of the 91 patients were not taking UDCA. Of these, 2 were AMA positive and had esophageal varices, 1 was AMA positive without esophageal varices, and the fourth was AMA negative and had esophageal varices. The reasons for not taking UDCA were side effects in 2 patients and poor compliance in the other 2 patients.Independent Predictors of Esophageal VaricesOf the 34 patients with esophageal varices, 15 (44%) had small varices and 18 (53%) had moderate/large varices. Variceal size was not available for 1 patient. Predictors of esophageal varices by univariate analysis are listed in Table 2. Continuous variables were categorized to maximize the average of sensitivity and specificity, and cut-off values were identified based on receiver operating characteristic curves. Table 3 shows the results of multivariate analysis using logistic regression including odds ratios and 95% confidence intervals. A Mayo risk score of 4.5 or greater and a platelet count of less than 140,000/mm3 were the only independent predictors of esophageal varices in this group of patients with PBC. Other variables were not shown to independently predict esophageal varices, including albumin level (P = .690), total bilirubin level (P = .723), model for end-stage liver disease score (P = .084), prothrombin time (P = .526), sex (P = .466), AMA status (P = .553), and age (P = .063).Table 2Predictors of Esophageal Varices by Univariate Analysis in 91 Patients With PBCVariablePatients with esophageal varices (n = 34)Patients without esophageal varices (n = 57)P valueFemales31 (91%)53 (93%)NSAge, y58.8 ± 1.856.2 ± 1.3NSSerum total bilirubin level (normal range, 0.0–1.0 mg/dL)2.8 ± 0.91.6 ± 0.6<.0001Serum albumin level (normal range, 3.5–5 mg/dL)3.5 ± 0.13.9 ± 0.1.003Serum alkaline phosphatase level (normal range, 35–129 U/L)336 ± 33336 ± 35NSSerum aspartate aminotransferase levels (normal range, 0–37 U/mL)89 ± 1071 ± 7.0NSSerum alanine aminotransferase levels (normal range, 0–41 U/mL)75 ± 1276 ± 8.6NSPlatelet count (normal range, 150–450 thou/cu mm)122 ± 11221 ± 14<.0001Prothrombin time (normal range, 10.5–13.5 s)12.5 ± 0.311.1 ± 0.3.0025Mayo risk score6.0 ± 0.34.5 ± 0.3.0003MELD score12 ± 18 ± 1.012NOTE. Values expressed as mean ± SE or number (%). Open table in a new tab Table 3Independent Predictors of Esophageal Varices by Multivariate Logistic Regression Analysis in 91 Patients With PBCPredictorP valueOR (95% CI)Mayo risk score ≥ 4.5.00910.6 (1.8–62)Platelet count < 140.0117.6 (1.6–37)OR, odds ratio; 95% CI, 95% confidence interval. Open table in a new tab By using contingency tables, we calculated the sensitivity, specificity, positive predictive values, and negative predictive values of independent predictors of esophageal varices. The results are shown in Table 4. A Mayo risk score of 4.5 or greater identified 92% of patients with PBC who had esophageal varices (sensitivity, 92%), whereas a score of less than 4.5 indicated a 93% likelihood of not having esophageal varices (negative predictive value, 93%). A platelet count of less than 140,000/mm3 indicated a 74% likelihood of having esophageal varices (positive predictive value, 74%). In this series, individuals with esophageal varices had platelet counts as high as 369,000/mm3. Combined, a Mayo risk score of 4.5 or greater and a platelet count of less than 140,000/mm3 had a positive predictive value of 80% and a negative predictive value of 96%.Table 4Test Characteristics of Independent Predictors of Esophageal Varices in 91 Patients With PBCSensitivitySpecificityPPVNPVAccuracyMayo risk score ≥ 4.592% (24/26)71% (25/35)71% (24/34)93% (25/27)80% (49/61)Platelet count < 14074% (23/31)82% (37/45)74% (23/31)82% (37/45)79% (60/76)PPV, positive predictive value; NPV, negative predictive value. Open table in a new tab The following model was generated when continuous values of platelet count and Mayo risk score were used to create a prediction formula: Y = −0.708 − 0.01* platelets + 0.5* Mayo risk score, where Y = ln (P−1/P), and P is the probability of varices.Cross-Validation of ResultsAmong 92 patients with PBC seen at University Hospitals of Cleveland and Metro Health Medical Center, 36 patients without a history of variceal bleeding and who had undergone EGD served as our validation group. The demographic and biochemical characteristics of these patients are shown in Table 1. Of the 36 patients, 24 underwent EGD for screening purposes, and 15 of the 24 had varices. The other 12 of 36 patients underwent EGD for the evaluation of unrelated symptoms, and, of these, 2 of 12 had esophageal varices. Fourteen of the 17 patients with esophageal varices had undergone liver biopsy, and 14% (2 of 14) had early fibrosis (Ludwig stages 1–2) on liver biopsy performed within 2 years of EGD.Table 5 shows the sensitivity, specificity, positive predictive value, and negative predictive value of a Mayo risk score of 4.5 or greater and a platelet count of less than 140,000 for the prediction of esophageal varices. In this set of patients, a Mayo risk score of 4.5 or greater had 100% sensitivity for predicting the presence of esophageal varices. In the absence of a Mayo risk score of 4.5 or greater, no esophageal varices were found. Logistic regression revealed that both variables independently predicted esophageal varices while controlling for the effect of other variables, including advanced histologic stage. The formula generated by multiple logistic regression analysis was tested in this independent set of patients. The mean probability of varices among patients who had esophageal varices was 0.75 (range, 0.44–0.97), whereas among patients without esophageal varices the mean probability was 0.28 (range, 0.03–0.71). The area under the receiver operating characteristic curve was 0.959 (Figure 1).Table 5Test Characteristics of Independent Predictors of Esophageal Varices in a Cross-Validation Set of 36 Patients With PBCSensitivitySpecificityPPVNPVAccuracyMayo risk score ≥ 4.5100% (17/17)63% (12/19)71% (17/24)100% (12/12)81% (29/36)Platelet count < 14059% (10/17)95% (18/19)91% (10/11)72% (18/25)78% (28/36)PPV, positive predictive value; NPV, negative predictive value. Open table in a new tab Within this cross-validation group, 3 of the 36 patients were not taking UDCA: 1 had side effects, another had not been started on the medication at the time of endoscopy, and a third patient discontinued use of UDCA because of a lack of liver enzyme improvement. These were all AMA-positive patients, and 2 of the patients had esophageal varices.DiscussionResults of this study suggest that the Mayo risk score and the platelet count can be helpful in predicting the presence of esophageal varices in patients with PBC. Specifically, a Mayo risk score of 4.5 or greater and a platelet count of less than 140,000/mm3 indicate a very high likelihood of esophageal varices. The prevalence of esophageal varices among patients with PBC who underwent EGD approached 50% in our institution. Because only half of our patients with PBC had an EGD, the high prevalence of varices likely represents a selection bias. Nevertheless, this pattern reflects common clinical practice in which only patients with known advanced liver disease are referred for endoscopy. This concept might change with the understanding that (1) a liver biopsy may not be necessary for the appropriate management of patients with PBC and (2) portal hypertension can occur in precirrhotic patients with PBC. Furthermore, the presence of esophageal varices may have prognostic implications.9Gores G.J. Wiesner R.H. Dickson E.R. et al.Prospective evaluation of esophageal varices in primary biliary cirrhosis: development, natural history, and influence on survival.Gastroenterology. 1989; 96: 1552-1559Abstract PubMed Google Scholar, 32Jones D. Walter R. Prince M. et al.The impact of portal hypertension in primary biliary cirrhosis.Gut. 2002; 50: A118PubMed Google Scholar Indeed, a recent Japanese study involving 84 patients with asymptomatic PBC showed that those with esophageal varices identified by screening EGD are more likely to progress to symptomatic disease within 3 years.33Takeshita E
Referência(s)