ACCURACY OF CARBOHYDRATE-DEFICIENT TRANSFERRIN IN THE DETECTION OF EXCESSIVE ALCOHOL CONSUMPTION: A SYSTEMATIC REVIEW

Hèlen Koch1,*, Gert-Jan Meerkerk3, Joost O. M. Zaat4, Maria F. Ham1, Rob J. P. M. Scholten2 and Willem J. J. Assendelft1,2,5

1 Department of General Practice, Division of Clinical Methods and Public Health and 2 Dutch Cochrane Centre, Academic Medical Center – University of Amsterdam, Amsterdam, 3 Addiction Research Institute, Rotterdam, 4 General Practitioner, Purmerend, 5 Department of General Practice, Leiden University Medical Center, Leiden, The Netherlands

* Author to whom correspondence should be addressed at: Academic Medical Center, University of Amsterdam, Division of Clinical Methods and Public Health, Department of General Practice, PO Box 22700, 1100 DE, Amsterdam, The Netherlands. Tel.: +31 20 566 3065; Fax: +31 20 566 9186, E-mail: h.koch{at}amc.uva.nl

(Received 26 July 2003; first review notified 4 September 2003; in revised form 25 November 2003; accepted 8 December 2003)


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Aims: Excessive alcohol consumption is a common problem in society and medical practice. There is a need for a diagnostic tool with both high sensitivity and specificity for the detection of excessive alcohol consumption in unselected medical populations. Therefore, we evaluated the diagnostic accuracy of carbohydrate-deficient transferrin (CDT) in the detection of excessive alcohol consumption. Methods: Computerised literature searches in Medline, Embase and Current Contents databases (01/1966–06/2003) and reference checking. Articles on the detection of excessive alcohol consumption reporting CDT levels and self-reported alcohol consumption as a reference test were selected (n = 101). Studies concerning treatment, relapse detection and traffic offenders were excluded. Quality assessment and data-extraction was done by two reviewers independently. Only studies scoring positive on core validity criteria by Lijmer were eligible for quantitative analysis (n = 29). Results: Only two CDT-assays (CDTect and CDTriTIA) were evaluated in more than two high validity studies fulfilling the criteria for inclusion in the statistical analysis. Sensitivity of CDTect (14 data points) ranged from 20 to 85%, whereas specificity ranged from 77 to 95%. A summary ROC curve was computed for CDTect. Sensitivity of CDTTriTIA (4 data points) ranged from 10 to 67%, and specificity ranged from 90 to 100%. No summary measure could be computed for CDTTriTIA. The heterogeneity of results could not be explained clinically. Conclusions: The validity of CDT as a diagnostic tool is still questionable. If the higher values for sensitivity that some studies report can be confirmed by others it is a useful diagnostic tool in unselected populations. However, more methodologically sound, comparable studies need to be performed before firm conclusions can be drawn.


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Excessive alcohol consumption is a common problem in society and medical practice. Due to definition differences, reliable prevalence figures of excessive alcohol consumption that allow international comparison are not available, but it is certainly not a rare problem in most countries. Excessive alcohol consumption may present in a broad variety of physical and psychosocial signs and symptoms, ranging from vague abdominal complaints to heart failure and Korsakoff's syndrome.

Once the physician suspects excessive alcohol consumption there are different methods to further confirm this possibility. First, there are several questionnaires and interviews focussing on the problems that might be caused by excessive alcohol consumption such as the Michigan Alcoholism Screening Test (MAST) (Selzer, 1971Go), its adaptations (Pokorny et al., 1972Go; Selzer et al., 1975Go; Kristenson and Trell, 1982Go), the CAGE questionnaire (Ewing, 1984Go) or the Alcohol Use Disorder Identification Test (AUDIT) (Saunders et al., 1993Go). Second, there are methods with diaries focussing on the quantification of alcohol consumption, for example quantity–frequency, time-period or time-line follow-back methods (Sobell et al., 1988Go; Webb et al., 1990Go). The difficulty with these specific questionnaires and quantification diaries is that in general people who drink alcohol tend to neglect or underestimate their alcohol consumption (Popham and Schmidt, 1981Go; Watson et al., 1984Go). Therefore, more ‘objective’ laboratory markers are often used, such as gamma-glutamyl-transferase (GGT) or mean corpuscular volume (MCV). However, applying the usual cut-off points for these tests, GGT turns out to have a low specificity whereas MCV shows a low sensitivity (Chick et al., 1981Go; Eckardt et al., 1981Go; Yersin et al., 1995Go). This may lead to unnecessary further testing or a gross misunderstanding with the patient.

Approximately 20 years ago another laboratory parameter, carbohydrate deficient transferrin (CDT), became available. This marker is a desialylated transferrin variant appearing in serum. It measures the accumulated effect of alcohol con-sumption, appearing after regular intake of 60–80 g of ethanol per day (approximately 7–10 units per day) for at least 1 week and normalizes slowly during abstinence (half-life = about 15 days) (Stibler, 1991Go). At its introduction, this test was supposed to be more specific and more sensitive than the former blood tests in the early studies (Stibler et al., 1978Go, 1979Go, 1980Go, 1986Go).

When a new test becomes available, it is often first studied in case-control studies, comparing the results in known cases (e.g. known alcoholics) with these in healthy controls (in many studies concerning CDT even teetotallers) (Lijmer et al., 2001Go). When promising results are found, the new test is subsequently evaluated in prospective studies, in which the new test is often first tested in selected samples and only later on in a broader spectrum of participants within different health care settings. Differences in design and patient characteristics influence test parameters like sensitivity, specificity and likelihood ratios (Lijmer et al., 1999Go). The final step in evaluating a new test is a thorough systematic review of the test performance in different circumstances and studies. Such a summary provides an estimate of the overall accuracy of the test in question, or, in case of heterogeneity, identifies relevant subgroups and their respective test parameters.

Others have already reviewed the CDT test in a narrative way (Stibler, 1991Go; Allen et al., 1994Go), sometimes in combination with a review of other laboratory markers (Rosman, 1992Go; Mihas and Tavassoli, 1992Go; Goldberg and Kapur, 1994Go; Potter, 1994Go; Yersin et al., 1995Go; Conigrave et al., 1995Go; Litten et al., 1995Go; Sillanaukee, 1996Go; Lesch and Walter, 1996Go; Allen et al., 1998Go; Rosalki, 1999Go). Recently, Salaspuro and Scouller used a more systematic approach (Salaspuro, 1999Go; Scouller et al., 2000Go). However, these reviews closed inclusion of studies at 1998 and Salaspuro did not assess the methodological quality of the included studies, which recently showed to be of utmost importance by Lijmer et al., who showed that methodological shortcomings lead to substantial overestimation of test parameters (Lijmer et al., 1999Go). Most of the available reviews concluded CDT being a more accurate laboratory marker for excessive alcohol consumption, more specific than GGT and more sensitive than MCV.

Because this conclusion may be of great importance for various health care fields we decided to conduct a state-of-the-art systematic review applying to the most recent methodological guidelines for conducting one, in order to reach a valid and applicable conclusion on the diagnostic accuracy of CDT in the detection of excessive alcohol consumption.


    MATERIALS AND METHODS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
This review was performed according to the most recent and widely accepted and used recommendations on (diagnostic) systematic reviews (Midgette et al., 1993Go; Irwig et al., 1994Go; Cochrane Working Group on Systematic Review of Screening and Diagnostic Tests, 1996Go; Devillé et al., 2001Go).

Search strategy
To identify articles on the subject a computerised search was performed, using MEDLINE (01/1966–06/2003), EMBASE (01/1988–06/2003) and Current Contents (04/1998–06/2003). We used the text words ‘CDT’, ‘carbohydrate deficient transferrin’ and ‘alcohol’ and subject headings regarding excessive alcohol consumption (drinking behaviour, alcohol-related disorders and alcoholic intoxication). The search was supplemented by reference checking. The complete search strategy is available upon request from the corresponding author.

Selection
We aimed to include articles that reported on CDT levels compared with a reference standard consisting of self-reported alcohol consumption (convertible in grams of alcohol per day). This reference standard [the ‘goldness’ (validity) of which is, of course, open to discussion] is often used in alcohol research. A more suitable and in comparative research widely used reference standard is to our opinion not available at present. Articles had to concern the detection of excessive alcohol consumption. As our aim was to evaluate the test in a general medical population, articles were excluded when they concerned treatment or relapse detection of excessive alcohol consumption and when handling traffic offenders. All retrieved references were first assessed on eligibility on basis of their titles, abstracts and keywords by two of the authors (H.K. and W.J.J.A.). A full text copy was retrieved when either presumed eligible or unclear and the same authors decided on final inclusion. Only articles written in languages mastered by all reviewers (English, German and Dutch) were included (Moher et al., 2000Go). Initial agreement, expressed as percentage and kappa, was computed for both steps of the inclusion (Brennan and Silman, 1992Go).

Quality assessment
Quality assessment of the included primary studies was done by two of the other authors (J.O.M.Z. and G.J.M.), with an adaptation of the quality assessment list for diagnostic studies of the Cochrane Methods Group on Systematic Reviews of Screening and Diagnostic Tests (Cochrane Working Group on Systematic Review of Screening and Diagnostic Tests, 1996Go). The used quality items are summarized in Table 1. The complete list is available upon request from the corresponding author. Special attention was paid to the subset of criteria, for which it has been proven that violation leads to overestimation of the accuracy of a diagnostic test (Lijmer et al., 1999Go), and to which we refer to as ‘Lijmer criteria’. Those criteria (marked * in Table 1) were: use of same reference standard (avoidance of selection and verification bias), blind interpretation of index test and reference standard results and appropriate spectrum of disease in the research population. The spectrum of disease was considered appropriate when the study concerned either consecutive patients or a selected sample of patient categories in which the components of the spectrum lie close to each other (e.g. moderate versus heavy drinkers). Spectrum of disease was called inappropriate when components of the spectrum were lying far apart from each other (e.g. known alcoholics versus teetotallers) or when the spectrum of disease was unclear.


View this table:
[in this window]
[in a new window]
 
Table 1. Items of study quality [operationalized in an adapted quality assessment list for diagnostic studies (Cochrane Working Group on Systematic Review of Screening and Diagnostic Tests, 1996Go)]

 
The quality assessors were blinded for authors, institutions and journal names (Jadad et al., 1996Go). Consensus was reached by discussing the differences. Measures of inter-observer agreement on item level were computed (Brennan and Silman, 1992Go).

Data extraction
Finally, only the studies scoring positive on all of the Lijmer criteria were considered eligible for the quantitative analysis. The quantitative data-extraction was performed by two of the authors (H.K. and M.F.H.) independently, reconstructing 2 x 2 contingency tables, extracting used cut-off points for the CDT test and, when necessary, converting the cut-off points of the reference standard into grams of alcohol per day. Differences were discussed. When no consensus could be reached a third author (W.J.J.A.) made the final decision.

Statistical analysis
An adapted decision tree based on that of Midgette et al. (1993)Go was used to decide which kind of meta-analytic method for summarising diagnostic test performances was most appropriate (Fig. 1). This algorithm takes into account the association between the extracted sensitivities and specificities, possible heterogeneity of the studies and identification of relevant subgroups. In this method the association between the sensitivities and specificities is expressed by Spearman's rho (R in Fig.1). When hetero-/homogeneity of studies needs to be tested, this is done by means of a Fisher's exact test.



View larger version (26K):
[in this window]
[in a new window]
 
Fig. 1. Schematic presentation of the statistical analysis strategy adapted from Midgette et al. (1993)Go.

 
In the analysis we only used the studies which compared the index test with a reference test with a cut-off range of 60–80 g alcohol per day.


    RESULTS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Search strategy and selection of studies
The results of the search and subsequent assessment of identified studies are summarized in Fig. 2. Of 480 hits, 212 studies (44%) were preliminarily selected and full text copies were obtained. Initial agreement on preliminary selection was 85%, kappa 0.7. Of this selection, 101 studies (21%) applied to the inclusion and exclusion criteria and were included for the quality assessment. Initial agreement between the authors on this inclusion was 84.8%, kappa 0.7. All disagreements on these two steps of inclusion could easily be solved with discussion of study details.



View larger version (22K):
[in this window]
[in a new window]
 
Fig. 2. Progress of the number of studies through the stages of the review.

 
Quality assessment
Initial agreement on item level between the quality assessors ranged from 60 to 99%. Kappas of this agreement are not reported here because these are hard to interpret, due to the high agreement-low kappa paradox (Feinstein and Cicchetti, 1990Go). Differences could probably partially be attributed to the different backgrounds of the assessors, one being a psychologist–alcohol addiction researcher and the other one being a general practitioner with methodological experience in systematic reviews. After discussion, however, consensus was easily reached in almost all cases. Twenty-nine studies (29%) met all Lijmer criteria. Of these 29, 26 contained sufficient quantitative information after data-extraction and, when necessary and possible, contacting the authors of the studies (six times).

Study characteristics
The characteristics of the 26 studies included in the analyses are listed in Table 2. Among the various studies a broad variety of test assays, cut-off points of index and reference tests, and patient populations was present. In some studies, the type of assay was unclear (Jaakkola et al., 1994Go; Lesch and Walter, 1996Go; Lott et al., 1998Go). In the remaining studies lots of different assays were evaluated (Legros et al.; Kaneko et al., 1994Go; Heil et al., 1994Go; Brathen et al., 2001Go). Actually, only CDTect- and CDTriTIA-assays were investigated in more than two studies.


View this table:
[in this window]
[in a new window]
 
Table 2. Characteristics of 26 high-validity studies concerning the accuracy of CDT for detecting excessive alcohol consumption

 
Test performance
The retrieved test parameters of the 26 studies are summarized in Table 3. Study results were highly heterogeneous. As stated before, only CDTect and CDTriTIA assays were investigated in more than two studies. The studies with these assays could therefore be further evaluated with the algorithm of Midgette. Only those studies comparing CDT with a cut-off point of the reference test of 60–80 g alcohol per day were further analysed. This strategy resulted in the quantitative analysis of 14 data-points from 11 studies concerning CDTect (marked * in Table 3) (Yersin et al., 1995Go; Stauber et al., 1995Go; Gronbaek et al., 1995Go; Sharpe et al., 1996Go; van Pelt, 1997Go; Meerkerk et al., 1999Go; Tonnesen et al., 1999Go; Sillanaukee et al., 2000Go; van Pelt et al., 2000Go; Brathen et al., 2001Go; Conigrave et al., 2002Go) and four data-points from three studies concerning CDTriTia (also marked * in Table 3) (Lesch et al., 1996Go; van Pelt, 1997Go; Mikkelsen et al., 1998Go).


View this table:
[in this window]
[in a new window]
 
Table 3. Performance of CDT in the detection of excessive alcohol consumption

 
CDTect
Sensitivity of the selected CDTect studies (marked * in Table 3) ranged from 20 to 84%, whereas specificity ranged from 77 to 95%. Sensitivity and specificity were inversely related (Spearman's rho: –0.08), implicating that a higher sensitivity in general was related with a lower specificity. The summary ROC curve (SROC) is shown in Fig. 3. The data pairs cluster in the left part of the graph, implicating that in the original studies cut-off points were applied that guaranteed a high specificity. The remainder of the SROC could not be plotted because of a lack of data points on lower specificities. Within the relatively small range of specificities, the sensitivities vary considerably. No relevant subgroups could be detected explaining this difference in sensitivities.



View larger version (14K):
[in this window]
[in a new window]
 
Fig. 3. Summary ROC curve of CDTect in the detection of excessive alcohol consumption.

 
CDTriTIA
Sensitivity of the selected CDTriTIA studies (marked * in Table 3) ranged from 10 to 67%, and specificity ranged from 90 to 100%. Sensitivity and specificity were not inversely related (Spearman's rho: +0.4). The sensitivities were hetero-geneous, both by visual inspection and by statistical testing (Fisher's exact test; P = 0.016). Consequently, summarisation of data was not possible.


    DISCUSSION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
More methodologically sound, comparable studies need to be performed before firm conclusions can be drawn concerning the accuracy of CDT in the detection of excessive alcohol consumption. Until then its usefulness, especially in settings where a high sensitivity is required, remains unclear because of large heterogeneity.

Our review illustrates that in its relatively short existence the CDT test, like many other tests, progressed through the various stages of test evaluation as mentioned in the introduction. The methodological more sound studies almost exclusively were published after 1993. Remarkably, reviews as early as 1991 (Stibler, 1991Go; Rosman, 1992Go; Mihas and Tavassoli, 1992Go), already concluded that CDT was a good test for detection of problematic alcohol consumption. The two recent, more systematic reviews were more careful. Salaspuro concluded that CDT was ‘slightly better’ than the conventional laboratory markers (Salaspuro, 1999Go). Scouller et al. already emphasized the need for ongoing and careful assessment of the evidence and for high quality research (Scouller et al., 2000Go).

Unresolved issues
This review highlights some important unresolved issues. In retrospect, the early reviews and editorials on this topic have actually been based on potentially flawed studies, as the studies applying to important validity items were only published relatively recently. We especially focussed on these methodologically more sound studies and in fact found that just few assays have been properly evaluated in more than two studies. The data on the assays that have been evaluated in multiple valid studies unfortunately do not allow a summary measure (a summary ROC curve with a sufficiently broad range of values or a pooled sensitivity/specificity). This heterogeneity brings us to the question which sensitivity or specificity can be applied for use in daily practice? Obviously, this question cannot be answered yet. We assumed that clinical heterogeneity could, at least partially, explain the statistical heterogeneity. Unfortunately, our quest for sensible subgroups revealed that most patient populations (like primary care patients, patients with vague complaints or high and low age groups) have not been evaluated in a sufficient number of studies. We also need more studies specifically on women, as there were only few of the remaining sound studies reporting on the accuracy of the CDT test in women.

Methodological considerations
We chose to analyse only those studies applying to certain core validity criteria, since Lijmer et al. empirically demonstrated the quantitative effect of these study design shortcomings on estimates of diagnostic accuracy (Lijmer et al., 1999Go). We wanted to get as close to the unbiased estimate of accuracy measures as possible. Applying the strict validity requirements resulted in a relatively small number of studies that still turned out to be very heterogeneous and difficult to summarize. Nevertheless, we did not want to make any concessions to these criteria.

Another interesting aspect is the choice of reference standard, in our review consisting of self-reported alcohol consumption (convertible in grams of alcohol per day). One can question the ‘goldness’ of this reference standard. Theoretically, the index test can be more accurate then the reference standard. One can imagine patients denying their drinking problems, underestimating the amount of alcohol they drink per day, and therefore scoring ‘negative’ on the reference standard. When their CDT level turns out to be above the cut-off value, these results would turn them from being true-positives (scoring positive on reference standard and index test) into being false-positives (scoring negative on the reference standard and positive on the index test). This shift would result in both a lower sensitivity and specificity of CDT. When subsequently the researchers keep the specificity of the index test high by increasing the cut-off point for the CDT test, this would even further lower the sensitivity, because of a shift of patients from true-positives to false-negatives. As a better reference standard is not available, we should be aware of the influence of an imperfect reference test on the found accuracy measures. Thus, perhaps the sensitivity of the CDT test is higher than we found.

Overall considerations
Much research has been done on the accuracy of CDT in detecting excessive alcohol consumption. Nevertheless, the variability of results between the high validity studies, especially the heterogeneity of sensitivity, could not be explained by presence of different subgroups or by methodological differences. Therefore, the validity of the CDT test as a diagnostic tool is still unclear, despite its already wide use and marketing. The sensitivity of the CDT tests in the lower range of the summary of values (Fig. 3) seems to be too low to be of practical clinical use. Consequently, the CDT test cannot be recommended in settings in which high test sensitivity is essential, like in unselected populations (low prevalence of the disease) or patients with unclear complaints (false-negative test results unwanted). For example, using the low sensitivity of 30% from the study of van Pelt (van Pelt et al., 2000Go), two out of three subjects with excessive alcohol consumption would be missed.

Recommendations
A summary of recommendations can be found in Table 4.


View this table:
[in this window]
[in a new window]
 
Table 4. Recommendations for future research

 
New studies in relatively unselected populations should be performed, applying to the Lijmer criteria in design, analysis and reporting in order to obtain valid estimates of test accuracy. Such studies should address study populations consisting of subjects with an appropriate, broad spectrum of disease, instead of just comparing cases (alcoholics) with controls (teetotallers). This implies a study in, for example, a primary care setting, or a population of patients with unclear complaints, in which the same reference standard is applied on all subjects in the research population and in which index test and reference standard results are blind interpreted. The importance of blind interpretation of the index test may, however, not be that important, as CDT values are computed by machine instead of by human beings and may therefore be less if not susceptible to bias. In contrast, the feasibility of a ‘blind’ quantification of alcohol consumption, the reference test, is hindered by the interaction between interviewer and subject that often occurs. Strict application of interviewing rules and instruction of subjects is required.

Furthermore, it is desirable that new test evaluations report on a broader variety of cut-off points of CDT than the present studies, not only on the high-specific ones. Because of the sub-optimal performance of the CDTect assay it might be interesting to keep on evaluating the other assays as well for comparison.

When more of these methodological sound comparable studies are performed, firmer conclusions can be drawn concerning the accuracy of CDT in the detection of excessive alcohol consumption. Until then it is unclear whether CDT does fulfil the expectation that it combines the sensitivity of GGT and the specificity of MCV in one assay.


    REFERENCES
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Allen, J., Litten, R. Z. and Lee, A. (1998) What you need to know: detecting alcohol problems in general medical practice. Singapore Medical Journal 39, 38–41.[Medline]

Allen, J. P., Litten, R. Z., Anton, R. F. and Cross, G. M. (1994) Carbohydrate-deficient transferrin as a measure of immoderate drinking: remaining issues. Alcoholism: Clinical and Experimental Research 18, 799–812.[ISI][Medline]

Brathen, G., Bjerve, K. S., Brodtkorb, E., Helde, G. and Bovim, G. (2001) Detection of alcohol abuse in neurological patients: variables of clinical relevance to the accuracy of the %CDT-TIA and CDTect methods. Alcoholism: Clinical and Experimental Research 25, 46–53.[ISI][Medline]

Brennan, P. and Silman, A. (1992) Statistical methods for assessing observer variability in clinical measures. British Medical Journal 304, 1491–1494.[ISI][Medline]

Chick, J., Kreitman, N. and Plant, M. (1981) Mean cell volume and gamma-glutamyl-transpeptidase as markers of drinking in working men. Lancet 1, 1249–1251.[Medline]

Cochrane Working Group on Systematic Review of Screening and Diagnostic Tests (1996). Recommended Methods, updated 6 June 1996. Available at: http://som.flinders.edu.au/fusa/cochrane

Conigrave, K. M., Degenhardt, L. J., Whitfield, J. B., Saunders, J. B., Helander, A. and Tabakoff, B. (2002) CDT, GGT, and AST as markers of alcohol use: the WHO/ISBRA collaborative project. Alcoholism: Clinical and Experimental Research 26, 332–339.[ISI][Medline]

Conigrave, K. M., Saunders, J. B. and Whitfield, J. B. (1995) Diagnostic tests for alcohol consumption. Alcohol and Alcoholism 30, 13–26.[Abstract]

Devillé, W., Buntinx, F., van der Windt, A. W. M., De Vet, H. C. W., Montori, V., Bezemer, P. D. and Bouter, L. M. (2001) Didactic guidelines for conducting systematic reviews of studies evaluating the accuracy of diagnostic tests. In The Evidence Base of Diagnosis, Knottnerus, J. A. ed., pp. 93–112. BMJ Publishing Group, London.

Eckardt, M. J., Ryback, R. S., Rawlings, R. R. and Graubard, B. I. (1981) Biochemical diagnosis of alcoholism. A test of the discriminating capabilities of gamma-glutamyl transpeptidase and mean corpuscular volume. Journal of the American Medical Association 246, 2707–2710.[Abstract]

Ewing, J. A. (1984) Detecting alcoholism. The CAGE questionnaire. Journal of the American Medical Association 252, 1905–1907.[Abstract]

Feinstein, A. R. and Cicchetti, D. V. (1990) High agreement but low kappa: I. The problems of two paradoxes. Journal of Clinical Epidemiology 43, 543–549.[ISI][Medline]

Goldberg, D. M. and Kapur, B. M. (1994) Enzymes and circulating proteins as markers of alcohol abuse. Clinica Chimica Acta 226, 191–209.[CrossRef][ISI][Medline]

Gronbaek, M., Henriksen, J. H. and Becker, U. (1995) Carbohydrate-deficient transferrin — a valid marker of alcoholism in population studies? Results from the Copenhagen City Heart Study. Alcoholism: Clinical and Experimental Research 19, 457–461.[ISI][Medline]

Heil, T., Spies, C., Bullmann, C., Neumann, T., Eyrich, K., Muller, C. and Rommelspacher, H. (1994) [The relevance of CDT (carbohydrate-deficient transferrin). Preoperative diagnosis of chronic alcohol abuse in intensive care patients following elective tumor resection]. Anaesthetist 43, 447–453. [In German.][CrossRef][ISI][Medline]

Irwig, L., Tosteson, A. N., Gatsonis, C., Lau, J., Colditz, G., Chalmers, T. C. and Mosteller, F. (1994) Guidelines for meta-analyses evaluating diagnostic tests. Annals of Internal Medicine 120, 667–676.[Abstract/Free Full Text]

Jaakkola, M., Sillanaukee, P., Lof, K., Koivula, T. and Nordback, I. (1994) Blood tests for detection of alcoholic cause of acute pancreatitis. Lancet 343, 1328–1329.[CrossRef][ISI][Medline]

Jadad, A. R., Moore, R. A., Carroll, D., Jenkinson, C., Reynolds, D. J. M., Gavaghan, D. J. and McQuay, H. J. (1996) Assessing the quality of reports of randomized clinical trials: is blinding necessary? Controlled Clinical Trials 17, 1–12.[CrossRef][ISI][Medline]

Kaneko, K., Fujimori, S., Yamanaka, H. and Akaoka, I. (1994) Effect of hypouricemic agents on serum carbohydrate-deficient transferrin in gouty patients. Advances in Experimental Medicine and Biology 370, 27–30.[Medline]

Kristenson, H. and Trell, E. (1982) Indicators of alcohol consumption: comparisons between a questionnaire (Mm-MAST), interviews and serum gamma-glutamyl transferase (GGT) in a health survey of middle-aged males. British Journal of Addiction 77, 297–304.[ISI][Medline]

Legros, F. J., Nuyens, V., Baudoux, M., Zouaoui, B. K., Ruelle, J.-L., Colicis, J., Cantraine, F. and Henry, J.-P. Use of capillary zone electrophoresis for differentiating excessive from moderate alcohol consumption. Clinical Chemistry 49, 440–449.

Lesch, O. M. and Walter, H. (1996) New ‘state’ markers for the detection of alcoholism. Alcohol and Alcoholism 31(Suppl. 1), 59–62.[ISI][Medline]

Lesch, O. M., Walter, H., Freitag, H., Heggli, D. E., Leitner, A., Mader, R., Neumeister, A., Passweg, V., Pusch, H., Semler, B., Sundrehagen, E. and Kasper, S. (1996) Carbohydrate-deficient transferrin as a screening marker for drinking in a general hospital population. Alcohol and Alcoholism 31, 249–256.[Abstract]

Lijmer, J. G., Mol, B. W., Heisterkamp, S., Bonsel, G. J., Prins, M. H., van der, M. and Bossuyt, P. M. (1999) Empirical evidence of design-related bias in studies of diagnostic tests. Journal of the American Medical Association 282, 1061–1066.[Abstract/Free Full Text]

Lijmer, J. G., Mol, B. W. J., Bonsel, G. J., Prins, M. H. and Bossuyt, P. M. M. (2001) Evaluation of diagnostic tests: from accuracy to outcome. In Evaluation of diagnostic tests: from accuracy to outcome, Thesis, University of Amsterdam, Amsterdam.

Litten, R. Z., Allen, J. P. and Fertig, J. B. (1995) Gamma-glutamyltranspeptidase and carbohydrate deficient transferrin: alternative measures of excessive alcohol consumption. Alcoholism: Clinical and Experimental Research 19, 1541–1546.[ISI][Medline]

Lott, J. A., Curtis, L. W., Thompson, A., Gechlik, G. A. and Rund, D. A. (1998) Reported alcohol consumption and the serum carbohydrate-deficient transferrin test in third-year medical students. Clinica Chimica Acta 276, 129–141.[CrossRef][ISI][Medline]

Meerkerk, G. J., Njoo, K. H., Bongers, I. M., Trienekens, P. and van Oers, J. A. (1999) Comparing the diagnostic accuracy of carbohydrate-deficient transferrin, gamma-glutamyltransferase, and mean cell volume in a general practice population. Alcoholism: Clinical and Experimental Research 23, 1052–1059.[ISI][Medline]

Midgette, A. S., Stukel, T. A. and Littenberg, B. (1993) A meta-analytic method for summarizing diagnostic test performances: receiver-operating-characteristic-summary point estimates. Medical Decision Making 13, 253–257.[ISI][Medline]

Mihas, A. A. and Tavassoli, M. (1992) Laboratory markers of ethanol intake and abuse: a critical appraisal. American Journal of the Medical Sciences 303, 415–428.[ISI][Medline]

Mikkelsen, I. M., Kanitz, R. D., Nilssen, O. and Huseby, N. E. (1998) Carbohydrate-deficient transferrin: Marker of actual alcohol consumption or chronic alcohol misuse? Alcohol and Alcoholism 33, 646–650.[Abstract]

Moher, D., Pham, B., Klassen, T. P., Schulz, K. F., Berlin, J. A., Jadad, A. R. and Liberati, A. (2000) What contributions do languages other than English make on the results of meta-analyses? Journal of Clinical Epidemiology 53, 964–972.[CrossRef][ISI][Medline]

Pokorny, A. D., Miller, B. A. and Kaplan, H. B. (1972) The brief MAST: a shortened version of the Michigan Alcoholism Screening Test. American Journal of Psychiatry 129, 342–345.[ISI][Medline]

Popham, R. E. and Schmidt, W. (1981) Words and deeds: the validity of self-report data on alcohol consumption. Journal of Studies on Alcohol 42, 355–368.[ISI][Medline]

Potter, B. J. (1994) Carbohydrate-deficient transferrin and the detection of alcohol abuse: a horse of a different color? Alcoholism: Clinical and Experimental Research 18, 774–777.[ISI][Medline]

Rosalki, S. B. (1999) Biochemical identification of alcohol abuse. International Journal of Clinical Practice 53, 138–139.[ISI][Medline]

Rosman, A. S. (1992) Utility and evaluation of biochemical markers of alcohol consumption. Journal of Substance Abuse 4, 277–297.[Medline]

Salaspuro, M. (1999) Carbohydrate-deficient transferrin as compared to other markers of alcoholism: a systematic review. Alcohol 19, 261–271.[CrossRef][ISI][Medline]

Saunders, J. B., Aasland, O. G., Babor, T. F., de, l. F. J. and Grant, M. (1993) Development of the Alcohol Use Disorders Identification Test (AUDIT): WHO Collaborative Project on Early Detection of Persons with Harmful Alcohol Consumption–II. Addiction 88, 791–804.[ISI][Medline]

Scouller, K., Conigrave, K. M., Macaskill, P., Irwig, L. and Whitfield, J. B. (2000) Should we use carbohydrate-deficient transferrin instead of gamma-glutamyltransferase for detecting problem drinkers? A systematic review and metaanalysis. Clinical Chemistry 46, 1894–1902.[Abstract/Free Full Text]

Selzer, M. L. (1971) The Michigan alcoholism screening test: the quest for a new diagnostic instrument. American Journal of Psychiatry 127, 1653–1658.[ISI][Medline]

Selzer, M. L., Vinokur, A. and van Rooijen, L. (1975) A self-administered Short Michigan Alcoholism Screening Test (SMAST). Journal of Studies on Alcohol 36, 117–126.[ISI][Medline]

Sharpe, P. C., McBride, R. and Archbold, G. P. R. (1996) Biochemical markers of alcohol abuse. QJM Monthly Journal of the Association of Physicians 89, 137–144.[ISI][Medline]

Sillanaukee, P. (1996) Laboratory markers of alcohol abuse. Alcohol and Alcoholism 31, 613–616.[Abstract]

Sillanaukee, P., Massot, N., Jousilahti, P., Vartiainen, E., Poikolainen, K., Olsson, U. and Alho, H. (2000) Enhanced clinical utility of gamma-CDT in a general population. Alcoholism: Clinical and Experimental Research 24, 1202–1206.[ISI][Medline]

Sobell, L. C., Sobell, M. B., Leo, G. I. and Cancilla, A. (1988) Reliability of a timeline method: assessing normal drinkers' reports of recent drinking and a comparative evaluation across several populations. British Journal of Addiction 83, 393–402.[ISI][Medline]

Stauber, R. E., Stepan, V., Trauner, M., Wilders-Truschnig, M., Leb, G., Krejsand G. J. (1995) Evaluation of carbohydrate-deficient transferrin for detection of alcohol abuse in patients with liver dysfunction. Alcohol and Alcoholism 30, 171–176.[Abstract]

Stibler, H. (1991) Carbohydrate-deficient transferrin in serum: a new marker of potentially harmful alcohol consumption reviewed. Clinical Chemistry 37, 2029–2037.[Abstract/Free Full Text]

Stibler, H., Allgulander, C., Borg, S. and Kjellin, K. G. (1978) Abnormal microheterogeneity of transferrin in serum and cerebrospinal fluid in alcoholism. Acta Medica Scandinavica 204, 49–56.[ISI][Medline]

Stibler, H., Borg, S. and Allgulander, C. (1979) Clinical significance of abnormal heterogeneity of transferrin in relation to alcohol consumption. Acta Medica Scandinavica 206, 275–281.[ISI][Medline]

Stibler, H., Borg, S. and Joustra, M. (1986) Micro anion exchange chromatography of carbohydrate-deficient transferrin in serum in relation to alcohol consumption (Swedish Patent 8400587–5). Alcoholism: Clinical and Experimental Research 10, 535–544.[ISI][Medline]

Stibler, H., Sydow, O. and Borg, S. (1980) Quantitative estimation of abnormal microheterogeneity of serum transferrin in alcoholics. Pharmacology, Biochemistry and Behaviour 13(Suppl. 1), 47–51.[CrossRef][ISI][Medline]

Tonnesen, H., Carstensen, M. and Maina, P. (1999) Is carbohydrate deficient transferrin a useful marker of harmful alcohol intake among surgical patients? European Journal of Surgery 165, 522–527.[CrossRef][ISI][Medline]

van Pelt, J. (1997) [Carbohydrate-deficient transferrin: a new biochemical marker for chronic excessive alcohol consumption]. Nederlands Tijdschrift voor Geneeskunde 141, 773–777. [In Dutch.][Medline]

van Pelt, J., Leusink, G. L., van Nierop, P. W. M. and Keyzer, J. J. (2000) Test characteristics of carbohydrate-deficient transferrin and gamma-glutamyltransferase in alcohol-using perimenopausal women. Alcoholism: Clinical and Experimental Research 24, 176–179.[ISI][Medline]

Watson, C. G., Tilleskjor, C., Hoodecheck-Schow, E. A., Pucel, J. and Jacobs, L. (1984) Do alcoholics give valid self-reports? Journal of Studies on Alcohol 45, 344–348.[ISI][Medline]

Webb, G. R., Redman, S., Sanson-Fisher, R. W. and Gibberd, R. W. (1990) Comparison of a quantity-frequency method and a diary method of measuring alcohol consumption. Journal of Studies on Alcohol 51, 271–277.[ISI][Medline]

Yersin, B., Nicolet, J. F., Dercrey, H., Burnier, M., van Melle, G. and Pecoud, A. (1995) Screening for excessive alcohol drinking. Comparative value of carbohydrate-deficient transferrin, gamma-glutamyltransferase, and mean corpuscular volume. Archives of Internal Medicine 155, 1907–1911.[Abstract]