Potential conflict of interest: Nothing to report.
Katharina Staufer was formerly affiliated with the Department of Hepatobiliary and Transplant Surgery, University Medical Center Hamburg-Eppendorf, Hamburg, Germany.
Optimal selection of liver transplant candidates and early detection of alcohol relapse after orthotopic liver transplantation (OLT) is necessary to improve long-term outcomes. In this study, urinary ethyl glucuronide (uEtG) was prospectively evaluated as a novel screening tool for alcohol detection in the transplant setting. Overall, 141 liver transplant candidates and recipients, visiting the outpatient clinic for a total of 308 times, were included. At each visit, the alcohol markers, uEtG, ethanol, methanol, and carbohydrate-deficient transferrin (CDT), as well as the state markers, alanine transaminase, aspartate transaminase, gamma glutamyl transpeptidase (GGT), and mean corpuscular volume (MCV), were determined, then compared to patients' self-reports on alcohol intake. Urinary EtG significantly increased the detection rate of alcohol consumption, compared to the other alcohol markers (P < 0.001). In 93% of patients and at 92.5% of visits with positive alcohol markers, alcohol intake was detected by uEtG and/or CDT. Sensitivity and specificity of uEtG were 89.3% and 98.9% and of CDT were 25% and 98.6%, respectively. Urinary EtG was the best independent predictor of alcohol consumption in univariate and multivariate analysis (positive predictive value: 89.3%; negative predictive value: 98.9%; odds ratio: 761.1; P < 0.001). It showed a superior prediction rate, when compared to established alcohol and state markers, as well as to the combination of CDT with MCV and GGT, assessed by net reclassification improvement (NRI) (NRI: 1.01, P < 0.001; NRI: 1.755, P < 0.001). Conclusion: uEtG is a sensitive, specific, and reliable marker for the detection of recent alcohol intake pre- and post-OLT. In combination with CDT, uEtG should be considered as a tool for routine alcohol screening within the transplant setting. (HEPATOLOGY 2011)
Relapse of alcohol consumption occurs in 11.5%-49% of patients undergoing orthotopic liver transplantation (OLT) for end-stage alcoholic liver disease (ALD).1, 2 Subsequently, graft dysfunction results in up to 17% and significantly reduces 10-year patient survival rates (45.1% versus 85.5%).3, 4
To facilitate the selection of patients with a low risk for alcohol relapse, most countries require a 6-month abstinence period immediately before transplantation.5, 6 Therefore, reliable screening tools for monitoring alcohol abstention are necessary. In addition to interviews, and the routine state markers, gamma glutamyl transpeptidase (GGT), aspartate transaminase (AST), alanine transaminase (ALT), and mean corpuscular volume (MCV), many transplant centers utilize the alcohol markers, carbohydrate-deficient transferrin (CDT) and ethanol (EtOH), as screening tests. However, these parameters have distinct limitations, particularly in patients with liver dysfunction or on multiple drug therapy.2, 7-12
CDT indicates continuous, excessive drinking (>50-80 g of ethanol/day) over a period of at least 1 week.9, 13, 14 However, in patients with end-stage liver disease, the specificity of CDT may be as low as 70%, and sensitivity ranges between 46% and 73%.7, 8, 15-17 A higher sensitivity and specificity of approximately 88% and 95%, respectively, has previously been achieved by combining CDT with MCV and GGT.7, 12, 13, 16, 17 These markers are also recommended for alcohol screening by the National Institute on Alcohol Abuse and Alcoholism of the National Institutes of Health (Bethesda, MD), if a patient is suspected to drink alcohol, but denies it.18 Yet, the value of this combination in patients with end-stage liver disease remains unclear.
In comparison to CDT, the measurement of EtOH only allows the assessment of recent alcohol consumption, because it can only be detected a few hours after alcohol consumption.19
Little experience on methanol (MeOH) as an alcohol-screening parameter is available in patients with liver dysfunction. In contrast to EtOH, MeOH usually remains detectable for approximately 2 days, but may accumulate in body fluids in cases of continuous heavy drinking, thereby also reflecting long-term alcohol intake.19-22 With a cutoff of 5 mg/L, as used in this study, a sensitivity and specificity of 70% and 98%, respectively, has previously been achieved.23
Urinary ethyl glucuronide (uEtG), an ethanol conjugate investigated in this study, exhibits several advantages. First, it can be detected for up to 80 hours after complete ethanol elimination from the body. This represents a longer detection window than for EtOH or MeOH.24-26 Second, it is detectable after the consumption of very small amounts of ethanol (≤5 g). Third, promising data on sensitivity and specificity were reported, ranging between 62.2% and 100% and 63.4% and 98%, respectively.27, 28 The value of uEtG as a screening parameter has lately been shown in outpatient treatment programs for alcohol and drug dependence (cutoff, 0.5 mg/L).29 In addition, it was recently implemented as a routine application in monitoring alcohol abstention for driving fitness (cutoff, <0.3 mg/L; or even <0.1 mg/L).21, 27
In this study, we evaluated the diagnostic and predictive value of uEtG, as compared to the alcohol markers, EtOH, MeOH, and CDT, as well as the state markers, ALT, AST, MCV, and GGT, in a cohort of transplant candidates and recipients.
Between January and September 2008, consecutive patients at the outpatient transplant clinic of the University Medical Center Hamburg-Eppendorf (Hamburg, Germany) were screened for alcohol consumption after written informed consent. Transplant candidates with underlying ALD or suspected alcohol consumption were included in the study. Additionally, transplant recipients with a history of alcohol abuse were tested at their yearly check-up visit or whenever alcohol consumption was suspected.
At each visit, patients were interviewed by a physician with regard to recent alcohol consumption, then blood and urine samples were collected. Urine samples were taken under the supervision of a nurse waiting in front of the bathroom door. When alcohol consumption was detected by any of the alcohol markers, patients were scheduled for another visit and confronted with their test results.
uEtG was screened by DRI-EtG-enzyme immunoassay (EIA; Thermo Fisher Scientific Inc., Passau, Germany). A cutoff of ≥0.5 mg/L was used to increase specificity and avoid false-positive test results.28, 30 In case of positive test results, the same urine samples were retested by liquid chromatography/tandem mass spectrometry (LC-MS/MS) for confirmation (Agilent Technologies 1100 Series high-performance liquid chromatography [HPLC]; Agilent Technologies, Waldbronn, Germany; and Waters Micromass Quattro Micro triple-quadrupole tandem mass spectrometer; Waters, Eschborn, Germany). For LC-MS/MS, selected reaction-monitoring equals of m/z 221 >75 and m/z 221 >85 (uEtG) or m/z 226 >75 and m/z 226 >85 (EtG-D5), respectively, were utilized. Lower limit of quantitation (LOQ) was 0.1 mg/L. All urine samples were stored at +4°C to +6°C and were analyzed within 2 days. To exclude any intentional dilution of urine samples, urinary creatinine was determined by Jaffé's method (Roche Diagnostics, Mannheim, Germany).31
Moreover, serum of patients was tested for the following alcohol markers: EtOH (head space gas chromatography/flame ionization detection, HS-GC/FID; Perkin-Elmer, Rodgau-Juegesheim, Germany), MeOH (HS-GC/FID; Perkin-Elmer), and CDT (%CDT by HPLC Reagent kit; Bio-Rad, Muenchen, Germany).10 Cutoffs were ≥0.1 g/kg, ≥5 mg/L, and >2.6%, respectively.
In addition to alcohol markers, the laboratory parameters, AST (cutoff, <35 U/L), ALT (cutoff, <50 U/L), GGT (cutoff, <55 U/L) (all NH4-heparine plasma), and MCV (cutoff, <94 fL; ethylenediaminetetraacetic acid/blood) were also assessed at each visit. The DeRitis ratio, defined as the AST/ALT ratio, was calculated. Values >1 were considered indicative of advanced ALD.32
Based on patients' self-reports after confrontation with test results, sensitivity, specificity, positive predictive values (PPVs), and negative predictive values (NPVs) were calculated. Correlation between alcohol and state markers was assessed by the calculation of Spearman's rank coefficients.
Univariate analysis of independent predictors of alcohol consumption was performed by computing diagnostic odds ratios (ORs), with a confidence interval (CI) of 95% for alcohol markers, state markers, and post-OLT status. Fisher's exact test was used to calculate P values. Multivariate analysis was done by excluding factors, which were insignificant in univariate analysis. To account for repeated measurements of the same patient, a repeated-measures logistic regression model using SPSS' GENLIN method (SPSS, Inc., Chicago, Illinois, USA) was utilized and included alcohol and state markers potentially predicting alcohol consumption. For each marker, the indicated cutoffs were used for calculation. In addition, results were adjusted for post-OLT status to model the presence or absence of liver cirrhosis, and to detect potentially divergent confession patterns, as possibly confounding factors.
Net reclassification improvement (NRI) was calculated to assess whether risk prediction for alcohol intake would improve when different alcohol markers were added to the measurement of state markers and to compare uEtG to established alcohol markers. NRI was computed using the predicted probabilities from the different repeated-measures logistic regression models.33
Statistical calculation was done by the Statistical Package for Social Sciences (SPSS for Windows, Rel. 15.0.1. 2006; SPSS, Inc.).
Of 141 patients (66.7% males; mean age: 53.5 years; range: 21-74), blood and urine samples were collected at 308 visits (median: 1.5; range: 1-10) to the outpatient clinic. In all patients, alcohol consumption caused or contributed to the underlying liver disease. In 6.5% of patients, chronic hepatitis C was a codiagnosis. One-hundred five (74.5%) patients were liver transplant candidates, whereas 33 (23.4%) patients were liver transplant recipients. In 3 (2.1%) cases, samples were collected before and after OLT. All liver transplant candidates had underlying liver cirrhosis, but no transplant recipient did.
Detection of Alcohol Consumption: Value of Alcohol and State Markers.
The patients' history of alcohol consumption, self-report on alcohol intake, and results of alcohol markers are summarized in Table 1. Importantly, all but 1 patient (patient 18; Table 1) denied alcohol consumption before alcohol testing. However, when confronted with the test results, 64.3% (18 of 28) of patients admitted recent or chronic alcohol consumption. In 10.7% (3 of 28) of patients, alcohol consumption remained unclear. This was caused by 1 patient not attending a follow-up appointment, and 2 patients refusing to make a statement. These 3 patients were withdrawn from statistical analysis.
Table 1. History of Alcohol Consumption and Results of Alcohol Marker Analysis
Overall, in 19.9% (28 of 141) of patients and at 13% (40 of 308) of visits, at least one alcohol marker was positive. It was noted that detection of alcohol consumption significantly increased by applying uEtG (P < 0.001). Urinary EtG was elevated in 71.4% (20 of 28) of patients and at 77.5% (31 of 40) of visits with any positive alcohol marker. In 50% (14 of 28) of patients, and at 62.5% (25 of 40) of visits with detected alcohol intake, uEtG was the only elevated marker. (Patients were defined as only positive for uEtG if another alcohol marker was neither positive at the same, nor at any other, visit.) However, in 28.6% (8 of 28) of patients, and at 22.5% (9 of 40) of visits, alcohol consumption was detected by MeOH and/or CDT, and not by uEtG. Ultimately, in 93% of patients at 92.5% of visits, alcohol consumption was detected by uEtG and/or CDT (Fig. 1).
Looking separately at pre- and post-OLT cohorts, 15.7% of transplant candidates were positive for at least one alcohol marker at 10.7% of visits (25 of 233; one marker: 19 of 233; two markers: 3 of 233; three markers: 3 of 233). In 58.8% (10 of 17) of patients, and at 68% (17 of 25) of visits, alcohol consumption was detected by uEtG. In 41.1% (7 of 17) of patients, and at 48% (12 of 25) of visits, uEtG was the only elevated marker. After transplantation, 30.6% of patients had positive alcohol markers at 20% of visits (15 of 75; one marker: 13 of 15; two markers: 1 of 15; three markers: 1 of 15). In 91% (10 of 11) of patients and at 93% (14 of 15) of visits, alcohol consumption was detected by uEtG. In 63.6% (7 of 11) of patients and at 53.3% (8 of 15) of post-OLT visits, uEtG was the only positive alcohol marker.
No significant correlation between elevated state markers and alcohol markers was verifiable by calculating Spearman's rank coefficients (data not shown). The frequency of elevated state markers in patients with and without positive alcohol markers is shown in Fig. 2.
Diagnostic Accuracy of Alcohol and State Markers.
Sensitivity, specificity, PPV, and NPV of alcohol and state markers are given in Table 2. Urinary EtG had, by far, the highest PPV and NPV, utilizing a cutoff of 0.5 mg/L. State markers, in particular GGT and DeRitis ratio, had high sensitivities, but low specificity and PPV.
Table 2. Prediction of Alcohol Consumption by State and Alcohol Markers
To exclude false-positive test results from the screening EIA, all urine samples with elevated uEtG were retested, and results were confirmed by LC-MS/MS (Table 1). To account for false-negative test results of uEtG by intentional dilution of the urine samples, urinary creatinine values were determined as described above. In 6.5% (20 of 308) of urine samples, urinary creatinine values were <20 mg/dL. Nevertheless, in 3 of these samples of 3 patients, uEtG was positive. In 3 further cases, MeOH was marginally elevated, whereas uEtG was negative (patients 23, 24, and 27; Table 1). Two of these patients denied alcohol intake (patients 24 and 27), whereas the third patient (patient 23), also positive for CDT at the same time, admitted alcohol consumption. The other 9 patients had no evidence of alcohol intake at any visit.
Applying a cutoff of 1.0 mg/L, instead of 0.5 mg/L, led to comparable results for specificity (98.9% versus 99.3%), NPV (98.9% versus 97.5%), and PPV (89.3% versus 91.3%), but to a decrease in sensitivity from 89.3% to 75%. Prediction of alcohol consumption within the multivariate analysis showed a decrease in OR from 761.1 to 412.5 (range: 80.6-2,111.5; CI: 95%) when using 1.0 mg/L as a cutoff (data not shown).
Prediction of Alcohol Consumption by Alcohol and State Markers.
Estimated ORs (95% CI) calculated by univariate analysis for state and alcohol markers are presented in Table 3. Calculation for EtOH was impossible, because only 1 patient had an elevated EtOH level. The alcohol markers, MeOH, CDT, and uEtG, predicted alcohol consumption with high significance, with uEtG having the most significant OR.
Table 3. Univariate Risk Estimation for Detection of Alcohol Consumption
ORs from the repeated-measures logistic regression models are shown in Fig. 3. In the first model (Fig. 3A), using state markers and the established alcohol markers, CDT and MeOH, positive CDT (OR: 13.2; P < 0.001), and post-OLT status (OR: 6.9; P = 0.002), significantly predicted alcohol consumption. After substituting CDT and MeOH (Fig. 3B) by uEtG, it was found that a positive uEtG value was the most important independent predictor of alcohol consumption, with an OR of 1,204 (P < 0.001). Alcohol intake was indicated by the presence of at least one positive alcohol marker and, therefore, was counted as an “event” within the calculation. All cases without positive alcohol markers were “nonevents.” This led to a separation problem in the statistical analysis when CDT, MeOH, and uEtG were compared in a single model. Despite this statistical challenge, uEtG showed a significantly higher prognostic value than either CDT or MeOH.
NRI was calculated to further evaluate the diagnostic benefit of uEtG (Table 4). NRI when comparing uEtG to the combination of CDT with MCV and GGT was 1.755 (P < 0.001) and, therefore, strongly supported the value of uEtG. Table 4A shows this reclassification using categorized probabilities. For 24 events (i.e., visits with reported alcohol consumption; upper triangle) and 261 nonevents (i.e., visits without alcohol intake; lower triangle), uEtG improved the predicted probability. In three events and three nonevents, results were better predicted by the established model without uEtG.
Table 4. Prediction of Alcohol Consumption by Net Reclassification Improvement (NRI)
NRI when comparing uEtG plus state markers (i.e., AST, GGT, and MCV included in analysis) with MeOH plus CDT plus state markers was 1.01 (P < 0.001). This also confirmed that uEtG outperforms the combination of MeOH and CDT in the detection of alcohol intake (Table 4B). Table 4C summarizes the reclassification when comparing uEtG plus state markers with state markers alone (NRI: 1.44; P < 0.001).
To account for possibly false statements, based on the patients' self-reports, NRI for only events was computed (0.862; P < 0.001). This probability reflects NRI of alcohol marker positive events only. It compares the model with state markers alone relative to the model with uEtG added. It is worth noting that the model with state and alcohol markers was significantly improved when using uEtG instead of CDT and MeOH, resulting in an NRI for events of 0.621 (P < 0.001).
In the transplant setting, it is of particular importance to detect alcohol consumption early and reliably, because transplant candidates unable to abstain from alcohol are taken off the waiting list. Additionally, transplant recipients who return to alcohol consumption should receive psychological and social support as early as possible to avoid irreversible graft damage.
In this cohort of liver transplant candidates and recipients, testing uEtG in addition to EtOH, MeOH, and CDT significantly increased (P < 0.001) the detection rate of alcohol intake. In total, alcohol consumption was detected in 19.9% of patients. In 71.4% of patients with positive alcohol markers, uEtG was positive, being the only elevated marker in half of the patients. In almost all patients (93%), alcohol consumption was detected by testing uEtG plus CDT. In this study, uEtG demonstrated a very high sensitivity and specificity, as well as a PPV of 89.3% and NPV of 98.9%, superior to the other alcohol markers, MeOH and CDT. In addition, uEtG was found to be a very strong independent predictor for alcohol consumption in the repeated-measures logistic regression analysis. Furthermore, the high NRI of 1.01, 1.44, and 1.755 (P < 0.001), respectively, supported the diagnostic value of uEtG, when compared to other alcohol and state markers. In addition, screening for uEtG is fast and inexpensive.
Notably, this study is the first to compare uEtG with the alcohol markers, EtOH, MeOH, and CDT, as well as with state markers in the transplant setting. With regard to the high detection rate of uEtG, the findings presented here are well in agreement with a pilot study conducted by Erim et al., who analyzed 18 liver transplant candidates with ALD. In that study, the results of breath alcohol testing, uEtG, and self-reports were compared. Although none of the patients admitted current alcohol consumption, and only one breath alcohol test was positive, as many as 49% of urine samples were positive for uEtG.34
Because patients with ALD and ongoing alcohol intake are not considered for OLT, specificity of an alcohol screening test is of particular concern. In a study conducted by Wurst et al., specificity of uEtG was only 68%, investigating 453 healthy subjects, drinkers, and abstinent patients. However, in this trial, LC-MS/MS, with a cutoff of 0.145 mg/L, was applied.27 On the other hand, in a publication of Boettcher et al., 400 urine samples were analyzed. By measuring uEtG with the DRI-EtG-EIA, and using a higher cutoff of 0.5 mg/L, a test specificity of 98% was achieved.28 Accordingly, this study used the DRI-EtG-EIA with a cutoff of 0.5 mg/L to avoid false-positive test results resulting from the EIA. In addition, all urine samples with elevated uEtG levels were confirmed by LC-MS/MS.35
An additional consideration is that ingestion of a very small amount of alcohol may lead to positive uEtG test results. Ethanol contained in mouthwash solutions, large amounts of fruit or vegetables, as well as ingestion of high amounts of baker's yeast may lead to elevated uEtG test results.36 Although the level of uEtG is usually expected to remain below the cutoff of 0.5 mg/L in such cases, a thorough history on eating habits is reasonable.
Another concern in liver transplant recipients is alcohol intake by ethanol-containing immunosuppressive or pain medications. In this study, neither medications nor food should have had a considerable impact on uEtG results, because all but 2 patients (patients 11 and 15; Table 1; both pre-OLT) positive for uEtG confessed alcohol consumption after confrontation with the test results.
Applying a cutoff of 1.0 mg/L, instead of 0.5 mg/L, for uEtG resulted in comparable results for specificity, PPV, and NPV, but a decrease of sensitivity from 89.3% to 75%. Notably, by using a cutoff of 1.0 mg/L, 4 patients with elevated uEtG and admitting alcohol consumption would have been missed.
Sensitivity of uEtG was previously found to vary between 62.2% and 100% in healthy subjects and patients with alcohol addiction.27, 28, 37 This study, including only patients with advanced liver cirrhosis and post–liver transplantation, achieved a high sensitivity of 89.3% by applying a cutoff of 0.5 mg/L. However, in 25% of patients, and at 22.5% of visits, alcohol consumption was detected by MeOH and/or CDT, and not by uEtG. False-negative uEtG results might have occurred because of bacterial degradation of alcohol and its metabolites in urine samples, despite the fact that urine samples were stored under chilled conditions to limit bacterial growth.23, 38
Furthermore, patients with positive MeOH levels should be expected to also test positive for uEtG. This is because MeOH and uEtG have a similar time window of detection. In this study cohort, in 3 cases, MeOH levels were elevated, whereas uEtG levels were not. In all of these cases, urinary creatinine values were considerably low, indicating a possible dilution of the sample. In 2 cases, MeOH was only marginally elevated, and alcohol consumption was not confessed (patients 24 and 27; Table 1). However, in the third case, in addition to MeOH, CDT was elevated, and the patient admitted alcohol consumption (patient 23; Table 1). Thus, the uEtG test result might be, indeed, false negative in this patient.
Based on the few data available, MeOH has a sensitivity of 70% in healthy subjects.23 In this cohort of pre- and post-OLT patients, it was only 22.2%. Nonetheless, specificity of MeOH, previously reported to be 98%, was here 99.3%. Taken together, according to the data presented here, the value of MeOH, in addition to uEtG, as an alcohol-screening parameter within the liver transplant setting is questionable.
In 17.9% (5 of 28) of patients with positive alcohol markers, CDT was the only elevated parameter at a single time point (patients 2, 15, 25, 26, and 28; Table 1). This indicates excessive drinking over a period of at least 1 week.9, 13 Yet, in 4 patients, CDT values were only slightly elevated (patients 2, 15, 26, and 28; Table 1). Additionally, none of these patients admitted alcohol consumption when confronted with test results by the physician. Although HPLC, used here for the detection of %CDT, is thought to virtually exclude false-positive results, some doubt about the possibility of false-positive CDT test results remains (PPV: 63.6%).10
The state markers, AST, ALT, GGT, and MCV, commonly used for the assessment of chronic alcohol abuse, have a relatively low specificity.39 This study confirms that, in patients with end-stage liver disease or liver transplant recipients, liver function parameters and MCV are not appropriate for reliable detection of alcohol consumption. In this study, PPV for all state markers was less than 16%. NPV, on the other hand, was over 90%, so that in the case of healthy liver function tests, alcohol consumption is unlikely.
It should also be taken into consideration that in this study, analysis of alcohol consumption was based on patients' confessions. The vast majority of patients admitted alcohol intake only after being confronted with the positive test results. Therefore, it is conceivable that the sensitivity of the tested alcohol markers, including uEtG, was overestimated. Though uEtG uncovered alcohol consumption in many pre- and post-OLT patients, it is possible that other patients remained undetected. More precise data on the accuracy of alcohol markers could be determined in a controlled study setting with test subjects consuming defined amounts of alcohol. However, in a transplant setting, including patients with intended long-term alcohol abstention, such a trial is ethically unfeasible.40
In conclusion, uEtG was found to be a reliable marker for the detection of alcohol consumption in transplant candidates and recipients. With a sensitivity, specificity, NPV, and PPV of over 89%, uEtG was superior to all other alcohol and state markers. However, despite the high specificity of the EIA-screening test used in this study, confirmation of positive samples by LC-MS/MS is still recommended until more data are available. In addition to uEtG, alcohol screening should also include the determination of CDT, because it reflects a different pattern of alcohol consumption and assesses excessive alcohol consumption over a longer time. Determination of MeOH and EtOH in addition to uEtG did not show substantial diagnostic benefit. Therefore, the assessment of uEtG plus CDT for screening liver transplant candidates and recipients is recommended.
The authors thank Jeffrey M. Sims for his help with editing this article.