|1||Smart, R.G., Ontario, Canada||General population survey, 703 drinkers, 18 and older||English, lifetime, personal interview||m/f, CAGE by question, and cut point 2||Results compared with alcohol intake, psychometric properties|
|2||Alvarez, J., Castile, Spain||General population survey, 2500, 53% drinkers, 14 and older||Spanish, lifetime, personal interview||m/f, CAGE by question, and cut point 2||Results compared with alcohol intake|
|3||Poulin, C., Canada||General population survey, 10530, 73% drinkers, 15 and older||English, past-year, telephone interview||CAGE pos. cut-point 2, independent risk factors for a positive result||alcohol intake, alcohol related problems|
|4||Tempier, R.T., Quebec, Canada||General population survey, 19724, 79%drinker||French, lifetime, questionnaire||CAGE by question||alcohol related problems, psychometric properties|
|5||Crowe, R.R., Iowa, U.S.A||Community sample 795 vs high risk sample 3435, 18 and older||English, lifetime and past year, personal interview||m/f, CAGE sensitivity, specificity, pos. pred. value for different cut-points, high risk vs community, alcohol abuse and dependence||DSM-III-R|
|6||Perdrix, A., Lausanne, Switzerland||Community sample, 416, 16 and older||French, lifetime, personal interview||m/f, CAGE pos. cut-point 2||Physicians diagnosis|
|7||Lairson, D.R., Houston, U.S.A.||Community sample, 687||English, lifetime, questionnaire||m/f, CAGE pos., specificity, cut-point 2, demographic and behavior variables||Addiction severity index, blood tests|
|8||Magruder, K., North Carolina, U.S.A.||Veteran outpatient-clinic, 915, only males||English, lifetime CAGE, personal interview||m, CAGE sensitivity, specificity for all cut-points lifetime vs current alcoholics, pos. pred. value||DSM-III-R|
|9||Buchsbaum, D., Virginia, U.S.A.||Outpatient medical practice, 821, 18 and older||English, lifetime, personal interview||sensitivity and specificity for different CAGE scores, likelihood ratios, ROC||DSM-III-R|
|10||Moret, V., Lausanne, Switzerland||Outpatient medical clinic, 270, 18 and older||French, lifetime, personal interview||m/f, age, CAGE pos., cut-point 2, characteristics of pos. population||MAST|
|11||Nystroem, M., Helsinki, Finland||First year university students, 2370, mean age 22, non-drinkers excluded||Finnish, lifetime, questionnaire||m/f, drinking habits, CAGE by questions and different cut-points||Alcohol intake, problems related to alcohol|
|12||Saunders, W., Glasgow, U.K.||Community sample, 3607||English, lifetime, personal interview||CAGE Sensitivity||Hospital records|
|13||Chan, A.W., Pristach, E.A.||General population (993)and general practice (390) samples||American, lifetime and past year, personal interview||m/f, drinking habits, sensitivity, specificity and Pos. pred. val.||DSM-III-R|
The CAGE Test shows good psychometric properties and suggests an unidimensional scale.
The rate of Cage positive drinkers (10.9 %) is similar to the percentage of drinkers who consume four or more standard drinks daily, derived from aggregate per capita consumption estimates. The authors conclude that this suggests that the CAGE cut-off of two positive answers identifies heavy drinkers consuming about four drinks a day. However, whether the two compared groups correlate with each other is not known.
Only 51.3 % of the sample consumed alcohol at least once a week. Of those 10.6% are Cage positive (cut-off 2), that is 5.4 % of the total sample.
Groups with different Cage scores (0 to 4) are compared to the mean alcohol intake of its members which shows a positive correlation. CAGE positive people have a 1.8 times higher alcohol intake than CAGE negative.
4.2 % of the sample report alcohol consumption over 80 grams/day and are therefore considered heavy drinkers. This number is compared to the 5.4 % of CAGE positive. Once again, it is not known if the two groups are identical.
The low amount of CAGE positive people is questioned considering the high alcohol intake of the Spanish population. Spanish cultural background that favors denial is given as a possible explanation.
That 49% of the population drink less than once a week is quite surprising compared to studies in other countries.
73.3% of the sample are current drinkers. To 5994 the Cage was administered. Of those 5.8 % screened positive. Regarding the whole sample, 3.4% are CAGE positive at a cut-point of two.
The same questionnaire inquired about harmful consequences occurring in the 12 months before the survey, arising from the respondents own use of alcohol:
|Problem area||CAGE negative
(% of respondents)
(% of respondents)
|Outlook on life||1.4||32.6|
|One or more areas||9.5||66.8|
The proportion of respondents reporting problems in one or more areas was 7 times greater among drinkers with a positive Cage result.
This study also shows that the higher the alcohol intake the more likely is a positive CAGE result. However it doesn't show the drinking habits of the Cage positive group.
Moreover male drinkers are 1.7 times more likely to have a positive result than females. But when male and female drinkers who had the same drinking pattern and other demographic characteristics were compared, there were no significant differences.
This study shows the good psychometric properties for the French version of the CAGE. Cronbach alpha is 0.70. Inter item correlation ranges from 0.23 to 0.49 (mean 0.37), the eye-opener question being responsible for the 0.23.
Additionally, questions about alcohol related problems are added which increases Cronbach alpha to 0.91.
The number of reported alcohol related problems are given but not compared to the CAGE.
This study compared a high-risk sample (35% alcoholics) to a community sample that is similar to general population (15%).
|Cut-point at 2||Lifetime||past year|
The lifetime prevalence of the low-risk sample is compared to the National Comorbidity Survey reporting 14.1% of alcoholism.
From all the all the different cut-off points, 2 performed the best regarding sensitivity and specificity. However it is noted that a lower cut-off in the low risk group may be more accurate, even though it produces a lower specificity (78%), because it allows detection of 85% of all alcoholics (vs 56% at a cut-off point at 2 in which case almost half of the alcoholics are missed.)
|CAGE cut-off 2||sensitivity||specificity||Pos. pred. value|
It is noted that the CAGE shows better sensitivity in the high-risk sample, although those values should be unaffected by the base rate. The same is noted for males. Possibly, these findings reflect greater severity of alcoholism in the high-risk group and in men.
Regarding alcohol abuse (2.1% of the whole sample) the CAGE performs less well than for alcohol dependence. However at a cut-point of 1 68% of the individuals were detected. This is a welcome finding because alcoholism may be more treatable in its early stage.
Overall, the authors conclude that the CAGE is a good screening-tool for alcoholism in a low-risk sample (such as general population) performing slightly better in male than in female. A cut-off point at 1 is suggested to detect most of the alcoholics, accepting some loss of specificity.
The results of the CAGE at a threshold of 2, administered by general practitioners were the follows:
This is surprisingly low for the studied population, especially in women. Moreover the group of alcoholics clinically diagnosed by the practitioners correlated weakly with the CAGE positive group.
This sample contains only males with a mean age of 54.4 years.
Performance of the CAGE for different cut-off points
Positive predictive values are calculated for different prevalence rates on the basis of the sensibility and specificity found in this study. This is questionable because other studies have shown that sensitivity and specificity change with the prevalence of alcoholism in a sample.
The authors conclude that the CAGE is an excellent screening tool for general clinics. However, prevalence of alcoholism in the studied sample is around 4% (positive MAST-test). The calculated pos. pred. value for this prevalence is reported to be only 0.2.
CAGE performance associated with specific CAGE scores:
Calculated posterior probabilities of being an alcoholic according to prior probability:
The authors suggest the use of likelihood ratios for different CASGE scores to interpret a patients risk for an alcohol problem rather than thresholds. However this is only possible when prior probability can be accurately estimated.
CAGE results at a threshold of 2:
Good correlation between MAST and CAGE (kappa=0.69) is reported.
The relative high results compared to other studies on alcoholism in the Swiss population are explained by the higher prevalence of alcoholism in outpatient clinics in the French part of Switzerland, not comparable to general practices or general population samples.
This Finnish study examines the drinking habits in first year university students.
Furthermore the CAGE showed good correlation with reported drinking amount.
Its performance in detecting heavy drinking was sensitivity of 50% and specificity of 84%.
This study is very unusual because the CAGE was administered to a general population that was not informed of the real purpose of the study. Without their permission records of psychiatric hospital in the studied region were compared to the test answers. Only 46% of the alcoholics in the studied sample as identified by their psychiatric treatments had a positive CAGE test. This performance is even more deceptive because it is known that 50% of hospital-treated alcoholics admitted to doorstep interviewer that they had psychiatric histories.
This study in the American general population compared the CAGE to DSM-III-R criteria.
38.3% of the population had a positive lifetime CAGE at a cut-off of two. 19.2% had a positive past-year CAGE.
Sensitivity and specificity are 77% and 85% respectively. Positive predictive value is 77.5. It concludes that the CAGE is a useful tool to screen for alcoholism or problem drinking in the general population.