A randomized controlled trial was conducted, with the HAPPY and HAPPY Plus provided for the control and intervention groups, respectively.

Participants were selected from employees of 10 companies that had agreed to collaborate in the study, who provided informed consent and fulfilled the following eligibility criteria: (1) aged over 20 years, (2) average daily alcohol intake of 2 drinks (20 g) for men and 1 drink (10 g) for women, and (3) no diagnosis of alcohol dependence.

Regarding the sample size, given an effect size of 0.33 for the BI [26] , a significance level of 5%, and calculation of significant differences with statistical power of 80%, the number of participants required for within-group measures and interactions in a two-way repeated measures ANOVA was 58 for each group. Using the same method and effect size, the required number of participants for between-subjects measures was 28 for each group. Assuming an attrition rate of 10% at 3 months, the target sample size was set at 128 (64 for each group) to produce the required sample size of 58 for each group.

Employees who, according to the results of workplace-provided health checkups required by the Industrial Safety and Health Act, were classified as people whose alcohol intake was at the heavy drinking level (540 ml or more per day) or people who already have a health disorder of γ-GT > 50 U/l were selected for participation. Recruitment was made through the companies by distributing flyers to each individual. After providing informed consent, participants were randomly assigned to the two groups.

Referring to the results of a recent field study that used focus group methodology to interview 12 male heavy drinkers [27] , the following revisions were made to the HAPPY:

1) A program title was created to improve recruitment. “Learning How to Have a Long-Term Relationship with Alcohol” was used as the title for the drinking moderation program, because 2 points had become clear. First, there was a strong resistance to reducing drinking (drinking provides an opportunity to socialize and build collegiality, and “being able to hold one’s liquor” is tied to the meaning of one’s existence, while drinking in moderation is believed to be connected to the disavowal of drinkers’ values). Second, there was a lack of foresight regarding long-term health (people do not think about being unable to drink because of health problems; it does not occur to them that their very existence could be threatened if they develop a health problem that makes it physically impossible to drink).

2) Changes were made to amplify participants’ perception of the threat to their health in order to increase behavior change rate. The absence of the perception of a threat to health results from a paucity of information and knowledge concerning alcohol-related health problems, with the benefits of drinking emphasized. Therefore, we revised the way in which information was provided, using the Health Belief Model [28] [29] , and improved the provision of medical information by instructing the therapist to present the information to participants interactively.

3) Stress management education was added. Because one of the reasons that participants drank alcohol to relieve stress, stress management education was integrated into the program to ensure that participants could learn to cope with stress via alternative means.

4) Measures were added to prevent dropout. We implemented interactive e-mail and telephone follow up between sessions, and integrated motivational interviewing skills into the program presentation, to strengthen communication regarding approval, encouragement, motivation, and sympathy.

5) The self-monitoring method was improved in order for participants to evaluate their drinking behavior objectively. We strengthened participants to set their goals on drinking behaviors and alcohol consumption and to monitor these achievements used a calendar-based diary.

6) Group dynamics were applied. Because customary drinking with work colleagues strengthens collegiality and provides a means via which to be oneself with others, we included group discussions regarding successful drinking (modeling), to increase self-efficacy.

A nurse facilitated this process. To assure the quality of this program, prior to the study, the nurse attended HAPPY training sessions and followed the manual prepared for this program.

As shown in Figure 1, participants were assigned to the control (HAPPY) or intervention (HAPPY Plus) group. The trained nurse was assigned as a facilitator for sessions 1 - 3 and follow up for both groups. One group consisted of about six participants. The durations of the sessions and between-session intervals were the same for both groups, with a period of one month between sessions 1 and 2 and two months between sessions 2 and 3 (Figure 1).

Session 1 consisted mainly of 1) an explanation regarding the fundamental volume of the workbook, alcohol consumption assessment, and setting moderate-drinking goals; 2) amplification of participants’ perception of the threat of alcohol-related health problems; and 3) participants’ recording of their alcohol consumption. To amplify the perception of threat, medical information presentation was interactive, and group discussion was included in the intervention group. Session 2 consisted mainly of 1) an explanation regarding the exercise volume of the workbook, alcohol consumption assessment (progress of implementation), and goal resetting; and 2) amplification of participants’ perception of the threat of alcohol-related health problems. Session 3 consisted of alcohol consumption assessment (progress of implementation), goal resetting, and a blood test.

First, to examine the effects of the HAPPY, a Friedman test was performed to determine the change in overall results over time for both groups, with the items measured at 3 time points for each group. A paired t test was performed to compare items measured at 2 time points (pre- and postintervention). To determine whether the HAPPY Plus yielded superior outcomes, a two-way repeated measures ANOVA or an unpaired t test was performed after determining normality. Participants’ achievement of moderate-drinking goals was expressed in terms of proportions of the participant group. A Friedman test was also used to examine changes in importance and confidence ratings over time for each group.

Because the study was performed office by office, participant recruitment and the study procedures were conducted between September 2012 and March 2015.

Ethical approval for the study was obtained from the Epidemiology Research Ethics Committee at the institution with where the authors were affiliated, and each office of the companies that collaborated in the study. The participants provided written informed consent to participation and the study was performed according to the Declaration of Helsinki.

Participants’ mean age was 46.1 ± 8.6 years (range: 25 - 60 years). Their mean alcohol consumption was 5.25 ± 3.23 drinks, and their mean AUDIT score was 14.0 ± 5.53 (Table 1). Statistically significant differences were not confirmed for control comparison.

Table 2 shows the changes in alcohol consumption over time and the physiological and psychological indicators for both groups combined and individually.

Daily alcohol consumption decreased significantly over time in the intervention and control groups and both groups combined (all, p < 0.001). However, according to the results of the two-way repeated measures ANOVA, consumption did not differ significantly between groups. Superior outcomes, therefore, were not observed for the HAPPY Plus with respect to the primary endpoint. AUDIT scores decreased significantly over time in the control group (p < 0.05) and both groups combined (p < 0.01), while scores did not change in the intervention group.

Regarding physiological indicators, BMI decreased significantly in the intervention group (p < 0.01) and both groups combined (p < 0.05). Systolic blood pressure also decreased significantly in the intervention group and both groups combined (all, p < 0.05). However, BMI and systolic blood pressure did not differ significantly between groups.

With respect to liver function, AST and ALT did not change subsequently to the intervention in both groups combined and individually (Table 3). However, γ-GT decreased significantly in the control group (p < 0.05) and both groups combined (p < 0.01), but did not change in the intervention group. In addition, γ-GT did not differ significantly between groups. Non-HDL and fasting glucose decreased very little; therefore, values did not differ significantly between or within groups.

Sex: chi-squared test; Psychological indicators: Mann-Whitney U test; Other: unpaired t test Abbreviations: γ-GT: γ-glutamyltranspeptidase; ALT: alanine transaminase; AST: aspartate transaminase; AUDIT: Alcohol Use Disorders Identification Test; non-HDL: non-high-density lipoprotein cholesterol; SD: standard deviation; n.s.: No Significant

Regarding γ-GT, a pre/postintervention comparison was performed for participants with baseline values exceeding the reference value (>50 U/l; Table 4). Large reductions, all of which were statistically significant, were observed in the intervention and control groups (all, p < 0.05) and both groups combined (p < 0.01). However, no significant difference was observed between groups.

Self-efficacy increased significantly in the control group (p < 0.05) and both groups combined (p < 0.01). However, scores in the intervention group did not change and did not differ significantly between groups. Scores in self-esteem were observed unchanged in both groups combined and individually (Table 2).

Participants’ goal achievement levels are shown in Table 5, and changes in their confidence levels over time and ratings for perception of the importance of preventing alcohol-related health problems are shown in Table 6. At both 1 and 3 months postintervention, relative to that of the control group, a higher proportion (60%) of participants in the intervention group had chosen the “goal almost achieved” response. Preintervention perception of the importance of preventing alcohol-related health problems was high in both groups, but no postintervention increases were observed. Confidence levels increased significantly in the intervention and control groups (all, p < 0.05). In addition, slight changes in goal achievement levels were observed in the intervention and control groups over time.

Table 7 shows the results of participants’ evaluations of the programs. The HAPPY Plus evaluation included the improvements applied to address the weak points of the HAPPY. Approximately 80% of participants rated the improvements as good or very good. In contrast, 32.5% rated the provision of information concerning stress

Abbreviations: γ-GT: γ-glutamyltranspeptidase; ALT: alanine transaminase; AST: aspartate transaminase; non-HDL: non-high-density lipoprotein cholesterol.

Abbreviations: γ-GT: γ-glutamyltranspeptidase.

management as “neither”. The HAPPY Plus revised alcohol consumption diary received a number of positive reviews.

As the participants’ baseline daily alcohol consumption and AUDIT scores indicated harmful use but did not fall within the alcohol dependence range; in addition, almost 40% of the participants already caused alcoholic liver disease, the study appears to have been successful in recruiting a sample in need of lifestyle modification. Collaborating with companies and the change in the program title facilitated the engagement process. These improvements appeared effective. In addition, the efforts to retain participants by e-mail and telephone follow up in the HAPPY Plus appeared successful, as 100% of the participants completed the study.

Since HAPPY Plus is a revision of the HAPPY, we examined effectiveness of both groups combined first. While significant changes occurred in the primary endpoint, average daily alcohol consumption, and secondary endpoints, AUDIT scores, BMI, systolic blood pressure, and γ-GT, were observed with both programs. The HAPPY, which provides feedback via the use of AUDIT scores to give participants with a clear understanding of their current alcohol use, relevant medical knowledge, and a tool of self-monitoring their consumption using diaries, exerted a direct effect on the reduction of alcohol consumption, showed the similar effect to that exerted by the BI [36] . As a secondary effect, the results confirmed a decrease in blood pressure, which was also suggested in the positive relationships between alcohol intake and hypertension shown in other studies [37] - [39] .

On the other hand, although alcohol intake is related to risk of glucose intolerance [40] and elevated fatty acid levels [41] , there were no improvements in these indicators. This may have occurred because there were only three participants whose results exceeded the reference values for diabetes. More important reason is that the program did not include diet and exercise education. This is explained in comparison with disease management programs which include comprehensive health risk behavior modification [35] and the metabolic syndrome lifestyle change education [42] , which proved improvement in indicators for not only alcohol consumption but also blood glucose, blood pressure, and lipids. Especially, Iyatomi et al. showed significant improvement for targeting high-risk drinkers whose AUDIT scores of at least 10 or consumed at least 21 drinks per week [42] . Therefore, we need to improve the programs; once the effects of moderate alcohol consumption have increased, the next step is to guide participants into programs including disease prevention programs that provide good dietary and exercise guidance or those that aim to improve overall health behavior.

In addition, scores for self-efficacy and self-esteem improved over time. This suggests that the intervention elements included in the HAPPY, such as empathy, goal setting and evaluation, could have been effective in providing a successful experience in drinking moderation.

Goal achievement was considerably higher in the HAPPY Plus, as further information concerning alcohol-re- lated health problems was provided, presentation was interactive, group dynamics were introduced, and telephone and e-mail follow up were added. Aside from this finding, the HAPPY Plus was not shown to be superior to the HAPPY.

A considerable reason is that there were control group participants whose data values were extreme, and improvements in these participants’ data exerted a strong influence on the results. Moreover, even in the HAPPY Plus, with behavior modification enhancement, the goal achievement rate was only 60%; notwithstanding the finding that there was a significant decrease in AUDIT scores in both groups combined. The average score did not fall below the dangerous consumption level. Further, despite a large and statistically significant reduction in γ-GT values in both groups combined, the average value did not decrease sufficiently to reach the normal range.

Another explanation for this could be that, even though the medium and method were altered, feedback, advice, goal setting, self-monitoring, and acquisition of medical knowledge were included in both programs. Moreover, as company employees, participants could be assumed to possess sufficient cognitive ability to adapt an approach to suit their needs and use the Internet and other resources to obtain information as required. In fact, some participants stated that the alcohol consumption diary required improvement and created their own spreadsheet programs. However, the diary used in the HAPPY Plus demonstrated some strengths. The calendar style exerted a self-monitoring effect, as participants were able to grasp the relationship between changes in their behavior and alcohol consumption, such as “the relationship between weekend dinner parties and alcohol consumption” and “setting alcohol-free days and its effects”, at a glance. While the introduction of a drinking diary was not shown to exert a significant effect in one study [43] , Hara et al. [44] described the effectiveness of moderate-drinking goals and a drinking diary. Nevertheless, with the finding that levels of achievement of behavioral goals were high in the HAPPY Plus, using a longer measurement period, it is conceivable that results could be reflected in improved physiological indicators and AUDIT scores.

In addition, participants’ evaluations of the stress management section of the program were low, and it did not exert a direct effect on controlling drinking behavior. Although entirely within the realm of conjecture, because this was a group intervention implemented in the workplace, it is possible that participants felt unable to discuss stress. Even though some studies have indicated that drinking behavior was associated with stress coping in men [45] , rather than drinking serving as a means of relieving stress, it could simply reflect a daily habit, as some participants value drinking for its own sake, and their drinking is not related to stress relief [27] . Therefore, stress management could be deleted for the group intervention.

Regarding the use of group interventions in the workplace, in recent years, Internet-based programs designed to assist individuals in reducing their drinking have multiplied in other countries, and there is evidence to support their effectiveness [46] . In particular, because such programs are easily accessible, they are useful for minors engaging in drinking illegally [47] . However, given that the development of web-based programs to reduce drinking has just begun in Japan, the use of group interventions should not be underestimated for several reasons. The biggest reason is that in Japan, drinking is often work related, and peer effect should not be underestimated. The results of this study suggested that convincing heavy drinkers to participate in and complete such programs requires an enforcement mechanism.

Both the HAPPY and HAPPY Plus were found to reduce daily alcohol consumption. In addition, to support participants with metabolic syndrome and respect to their dietary problems, the revised version of HAPPY that includes improvement of overall health behaviors is required. To improve program participation and completion rates, it was important to direct appeal for participation to companies, establish work-based groups, and conduct e-mail and telephone follow-up between sessions. Finally, since goal achievement rate was high in HAPPY Plus, it might be necessary to observe changes in drinking behavior over a period of at least a year rather than relying on surrogate indicators.