OUP user menu


DOI: http://dx.doi.org/10.1093/alcalc/agh095 548-551 First published online: 20 September 2004


Aims: In 1986, 338 men and women attending a general population screening study were identified as at-risk alcohol drinkers and randomly assigned into three groups. Two groups received slightly different, brief interventions; a third group served as control. After 1 year there was a 50% reduction in alcohol intake in the intervention groups and a 20% increase in the control group. Controls then received advice to reduce their drinking. This study evaluates outcomes 9 years after these interventions. Methods: In a survey in the same city in 1995 (over 27 000 participants), 247 subjects (73.1%) from the 1986 study, were re-assessed. Serum gamma-glutamyltransferase (GGT) was examined and compared with values in 1986. A ‘pseudo-control’ group was established to compare ‘treated’ and ‘untreated’. Results: After 9 years, the original study groups displayed significant mean reduction in GGT. The reductions achieved in the three groups did not significantly differ from each other. However, the reductions were significantly greater than that in the ‘pseudo-control’ group. Conclusion: The impact of brief intervention appears to be long lasting. At 9 years follow-up, the at-risk drinkers displayed GGT values close to that of the background population.

(Received 16 April 2004; first review notified 17 May 2004; in revised form 18 June 2004; accepted 24 August 2004)


Reviews have highlighted the importance of early identification of alcohol risk drinkers and the effectiveness of a subsequent brief intervention (Saunders and Conigrave, 1990; Bien et al., 1993). A WHO group of alcohol researchers who have developed instruments for identification of risk drinking have described the principles of an effective brief intervention, and proposed a nation-wide implementation of this approach in general practice (Anderson, 1985; Saunders et al., 1987, 1992). Numerous studies have described the short-term effect of brief intervention (6–12-month follow-up), whereas the long-term effect has only been evaluated in a few studies (Kristenson et al., 1983; Fleming et al., 2002; Wutzke et al., 2002).

In the Third Tromsø Study (1986), a community based, comprehensive health survey aimed at coronary heart disease and general health, with more than 20 000 participants, we had the opportunity to conduct a randomized controlled trial of brief intervention in 338 (non-dependent) subjects, identified by increased GGT-levels and an alcohol intake of at least 2–3 times a week, and defined as ‘at-risk alcohol drinkers’ (Nilssen, 1991). One group received a ‘minor intervention’ of ∼10 min (during which they were asked to consider possible reasons for their elevated GGT), and the other a 15 min intervention (during which they were counselled about how to reduce their GGT levels by proposing alcohol-free days, changing to lower alcohol beverages, etc.) with subsequent monthly visits until their GGT values were normalized (‘major intervention’). A third group, the control-group, remained ‘untouched’ until follow-up 1 year later when they also were subject to brief intervention. Details about identification, intervention and results are described elsewhere (Nilssen, 1991).

At 1-year follow-up we demonstrated a significant reduction in alcohol intake in both intervention groups, and a corresponding reduction in GGT-levels. The control group, however, showed a significant increase both in daily intake and in GGT level.

The current report, somewhat delayed for personal reasons, addresses the question of whether the observed intervention effect only had a ‘short term effect’ or whether it persisted over time.

In the Fourth Tromsö Study (1995) with more than 27 000 subjects, we were able to re-examine 247 of those who were subject to intervention in 1986, and thereby evaluate the long term effect of our intervention 9 years earlier.


In the Fourth Tromsö Study (1995) all inhabitants of Tromsö older than 24 years were invited to a health screening which, similar to the previous Tromsö Study, also had its main focus on coronary heart disease. In all, 27 161 subjects (77% of the eligible population) were screened. A protocol similar to that used in the previous surveys in this population (Stensland-Bugge et al., 1997) and in the Norwegian county studies (Bjartveit, 1979), was followed. The participants answered a six-page questionnaire, including several questions on alcohol use. All participants had blood-samples drawn for measurements of several blood parameters, among them GGT. As the recommended methods for measurements of GGT were slightly different in the two studies (recommendations from the Scandinavian Enzymes Committee; Committee on Enzymes of Scandinavian Society, 1976) in 1986 versus recommendations from the International Federation of Clinical Chemistry (Shaw et al., 1983 in 1995), all 1995 values of GGT were converted to 1986 values for comparison (Strømme et al., 1990).

Of the 338 subjects who, in 1986, were identified as at-risk drinkers and subject to brief alcohol intervention, 247 attended the health screening in 1995. They underwent the same examinations as the larger study population. They were not subject to individual interview on alcohol use. As the control group had received brief intervention at the 1-year follow up in 1987, they were no longer to be used as a control group for the 9-year study. The criteria for inclusion in 1986 were GGT values between 50 and 200 U/l for men, 45–200 U/l for women, and an alcohol intake of at least 2–3 times a week. We therefore established a pseudo-control group of all subjects who, in 1986, had GGT values slightly below inclusion level (i.e. between 45 and 49 U/l for men; 40–44 U/l for women) and a reported alcohol intake of at least 2–3 times a week. In all, 95 subjects met the criteria for pseudo-controls. Of those, 62 subjects were re-examined in 1995, thus constituting our new control group (pseudo-controls).

Statistical analyses were performed using the SAS statistical software package (SAS Institute, 1988). Analyses of variance were used for testing of differences in GGT reduction within and between the groups. The differences were adjusted for mean levels of GGT values in 1986 and 1995. As the basic GGT values (1986 values) are correlated with the change in GGT, we used the means of both measurements (1986 values + 1995 values/2) for the adjustments. These means are uncorrelated with the GGT changes, and also give a more stable standard error (SE).


Of the 338 at-risk drinkers from the brief intervention study in 1986, 247 attended the survey in 1995 (Table 1). According to the National Population Register, 13 subjects had died (equally distributed in the three groups) and 30 had moved from Tromsö between 1986 and 1995. Thus, the attendance rate at follow-up was 73.1%. Of the 95 subjects who met the criteria for inclusion in the pseudo-control group, 62 subjects met the criteria in 1995, giving a follow-up rate of 65.3%. There was no difference between the two follow-up rates (Fisher's two-tailed P = 0.158, chi-squared P = 0.137).

View this table:
Table 1.

Study population at the follow-up in 1995

Examined 1986Moved or dead Total (dead)Follow-up 1995 Attended
Control-group11215 (4)83 (74.1%)
Minor intervention11313 (5)87 (77.0%)
Major intervention11315 (4)77 (68.1%)
Total33843 (13)247 (73.1%)

There were no substantial differences in age, blood pressure, heart rate or lipids between the different groups (Table 2) but, as already demonstrated in 1986, lipids (with exception for HDL cholesterol) and blood pressure were significantly lower in the background population (6.05 mmol/l) than they were in the alcohol groups. The major intervention group contained more men than did the other groups, and the intervention group also smoked more and drank more coffee. The pseudo-control group displayed means within the range of the other groups for all variables except triglycerides, which were significantly lower (P < 0.001).

View this table:
Table 2.

Age- and sex-adjusted characteristics in groups at 9-year follow-up (1995)

Pseudo control (n = 62) (SD)Control group (n = 83) (SD)Minor intervention (n = 87) (SD)Major intervention (n = 77) (SD)
Age (years)*41.2 (11.2)42.9 (10.2)42.0 (9.9)41.2 (9.6)
Men (%)85.579.588.594.8
Cholesterol (mmol/l)6.57 (1.3)6.44 (1.3)6.50 (1.4)6.69 (1.4)
Triglyceride (mmol/l)1.76 (1.5)2.14 (1.5)2.20 (1.6)2.04 (1.6)
GGT (U/l)43.5 (23.8)71.4 (54.4)62.3 (43.1)54.8 (29.9)
HDL (mmol/l)1.60 (0.5)1.55 (0.5)1.48 (0.5)1.56 (0.5)
Systolic BP (mmHg)139.9 (19.7)140.5 (19.6)141.3 (20.9)138.4 (21.4)
Diastolic BP (mmHg)82.7 (13.2)82.8 (13.2)82.8 (14.0)82.3 (14.4)
Heart rate/min74.6 (16.2)75.8 (16.2)76.5 (17.2)77.4 (17.6)
Coffee (number of cups)5.1 (3.9)4.5 (3.9)4.2 (4.2)5.4 (4.3)
Smokers (%)41.535.941.047.0
  • * Age in 1986.

Table 3 shows crude and adjusted values for GGT changes between 1986 and 1995 for the trial groups and the pseudo-control group. Only the major and the minor intervention groups displayed reductions significantly different from 0 when crude values were compared (P = 0.0009 and 0.0011, respectively), but there were no significant differences in the GGT change between the groups (P = 0.2176). In analysis of variance with adjustment for mean level of GGT in 1986 and in 1995, all three groups who had received treatment displayed significant GGT reduction at follow up (P < 0.05), but no difference was found between the groups (P = 0.439). When the pseudo-control group was also included in the covariance analysis (column 3), a highly significant difference was seen between the pseudo-controls and the trial groups (P < 0.0001). Subjects in the pseudo-control group had increased their GGT to a level that was almost significant (P = 0.06), whereas the treated groups displayed significant reductions (P ≤ 0.0024). Inclusion of the non-attendants in the analysis (giving them the same GGT value as in 1986) resulted in an increased difference between the pseudo-controls and the trial groups (P < 0.0001; F = 9.49; d.f. = 4/429). This difference was further increased when adjustment for the average age-related change in GGT was undertaken (F = 14.34; P < 0.0001).

View this table:
Table 3.

Analysis of variance with crude and adjusted* mean difference in GGT between 1986 and 1995 for different groups

Crude values in all groupsAdjusted* values in the trial groupsAdjusted* values in pseudo and trial groups
Pseudo control group**62−4.23.06−1.370.17578.974.771.880.0610
Control group83−6.25.43−1.140.2576−9.594.32−2.220.0272−13.184.00−3.300.0011
Minor intervention87−15.64.61−3.380.0011−16.274.19−3.880.0001−19.633.86−5.09<0.0001
Major intervention77−14.04.06−3.450.0009−9.554.50−2.120.0346−12.494.08−3.060.0024
Analysis of varianceBetween all groupsBetween the trial groupsBetween pseudo and trial groups
Degrees of freedom (d.f.)3/3053/2444/305
P-value0.21760.4393< 0.0001
  • * Adjusted for mean levels of GGT in 1986 and 1995.

  • ** Comprise subjects just below inclusion criteria for intervention.


This study explores the effect of brief alcohol intervention after a period of 9 years. The effect is described in terms of change in GGT-levels.

Follow-up population

Even though the attendance rate at follow-up was 73.1% (n = 247), 83.7% of those invited (exclusion of participants who had died or who had moved) were re-examined in 1995. This compares favourably with other studies. There had been no contact with the at-risk drinking groups between the 1- and 9-year follow-ups, no special invitation was sent in 1995 and no effort was made to contact them if they did not attend the screening. They were re-assessed in the large health screening for coronary heart disease but were not subject to personal interviews. For the non-attendants, we have reason to believe that many of these subjects were temporarily absent from Tromsö for educational or work-related reasons during the screening period.

Limitations of the study

In 1986 our regional ethical committee insisted that our control group should receive intervention at the 1-year follow-up, thus removing the possibility of having the same control group over time, thereby causing a potential weakness in the study design. A pseudo-control group was therefore established at the 9-year follow-up, consisting of subjects with GGT levels just below the inclusion criteria but with an intake frequency equal to that of the study population. In this study we therefore consider the original control group as a treated group.

For methodical reasons, only one group was asked to quantify their alcohol intake at the intervention. As the participants were randomly assigned into different groups, we assumed that mean alcohol consumption in this group represented a fair estimate of alcohol intake in the two other groups. This assumption is clearly questionable. Hence we feel unable to present data based on volume of alcohol consumed, a fact that impairs the strength in this study.

Missing quantification in individual drinking frequency at intervention represent another potential weakness. The way the categories for inclusion were defined in 1986 (2–3 times a week, daily or about daily) did not allow for an exact determination of frequency. Even though information on frequency was more accurate in 1995, our possibility to estimate individual change over time is limited.

Comparison with other studies

Effect from the intervention was still visible after 9 years. The effect was not only seen in the initial intervention groups but also in the control group, who received brief intervention at the 1-year follow-up. Similar results were found in the Malmö Study (Kristenson et al., 1983), one of the studies with more than 1-year follow-up. They also reported significant reductions in GGT level after 4 years, both in the intervention and in the control groups, but effect on level of alcohol use was not reported. In contrast to the Tromsö study, however, their participants were offered continuing contact throughout the follow-up period. When the GGT value had stabilized at an acceptable level, the frequency of therapeutic contacts was reduced but not stopped. Their control group also received intervention during the follow-up, as they were informed that they had an impaired liver test and were asked to restrict their alcohol intake.

In a randomized control study on brief intervention from Wisconsin (Fleming et al., 2002), the treatment group exhibited significant reductions in 7-day alcohol use at 48-month follow-up. They also reported significant fewer episodes of binge drinking and lower frequency of excessive drinking.

An Australian study found no effect of brief intervention after 10 years (Wutzke et al., 2002). The median consumption in the intervention and the control groups was similar, and mean reduction in consumption from baseline to follow-up was larger in the intervention group than in the control group, but this latter difference was not statistically significant. Although the Tromsö and the Australian studies included only non-dependent alcohol consumers, participants were recruited differently. In the Australian study the participants had a higher consumption level at baseline, different criteria were used for the definition of unsafe drinking and elevated GGT levels, and the criteria for inclusion and exclusion were also somewhat different, all of which might limit the comparability with our study.

Interpretation of results

In 1995 our participants were re-identified in the larger population study, which primarily focused on coronary heart disease, with more than 27 000 participators. An eager-to-please bias does not seems likely as the participants were not informed about our follow-up.

We had expected an increase in GGT levels, mainly for two reasons. First, according to the distribution of GGT in a normal population (Nilssen et al., 1990) an age-related 10% increase in men and a 20% increase in women would have been expected after 9 years. Second, as argued in the theoretical basis for early identification and early intervention, at-risk drinkers tend to increase their alcohol intake (and thereby their GGT level) over time. Nevertheless, a reduction in GGT level for the total risk drinking population was seen after 9 years (Table 3). The reduction (unadjusted) was strongest in the major and minor intervention groups (20.3 and 20%, respectively), although a somewhat smaller reduction for the control group (8%) was also observed. Regression towards the mean may explain parts of this reduction, but as the reduction in GGT was significantly larger for the treated groups than for the pseudo-control group, we suggest this might be interpreted as an effect of the intervention in 1986. The alcohol consumption trend, both in Tromsö and in the society at large, slowly increased during this period (from ∼4.2 to 5.7 l annual intake per capita). Increased awareness of health hazards as a result of repeated health investigations is possible but not very likely, as the large, background population displayed a significant increased (3.9 %) in GGT level during the same period (data not shown).

In analysis of variance (Table 3), after adjustment for mean levels of GGT in 1986 and 1995, no difference in GGT reduction between the trial groups was seen. This fits well with the findings at the 1-year follow-up. In 1986, we concluded that the minor intervention proved just as effective in reducing alcohol intake as the major intervention. This conclusion also seems valid after 9 years. A 15-min consultation with subsequent monthly control of GGT until normalization (major intervention) resulted in an equal GGT reduction as a single 10-min brief of information about the risk status without any further control (minor intervention).

The trial groups, however, displayed significantly greater GGT reduction than did the pseudo-control group (Table 3, column 3), indicating a beneficial effect of the intervention.

If, for the benefit of an analysis by intention-to-treat, all non-attendants were given the same GGT value in 1995 as in 1986 and included in the variance analysis, the difference between the trial groups and the pseudo-control group would have been further increased. An analysis of this kind must assume that the non-attendants did not change their alcohol intake, an assumption that can not be justified from our data.

For methodical reasons, only one group (the major intervention group) quantified their alcohol intake at the time of the intervention, reporting an average intake of 32.6 g pure alcohol per day. At the 1-year follow-up, the original control group reported an intake of 39.2 g (before receiving intervention). After 9 years the three intervention groups reported intake between 8 and 10 g pure alcohol per day (less than 1 alcohol unit), which is far below international recommendations for a safe level of intake (Bondy et al., 1999). In addition, frequency of ‘any alcohol intake’ was reported as being reduced after 9 years. While the criteria for enrolment in 1986 was intake of alcohol ‘2–3 times per week’ or ‘daily or about daily intake’, average intake in 1995 was less than 1.4 drinking episodes per week, thus indicating a reduction in drinking frequency by more than 50%. Both the reduction in volume and frequency of alcohol intake lends support to the demonstrated GGT reduction, thus completing the picture of the efficacy of brief intervention.

In conclusion, the view that the effects from brief intervention in at-risk alcohol drinkers only represent short-term effects was not confirmed in this study. On the contrary, at-risk drinkers seem to respond beneficially to brief intervention.


View Abstract