A Controlled Trial of an Internet-Based Intervention Program for Cannabis Users

The 50-day counseling program QTS comprises four consecutive phases: (1) The registration procedure mainly consists of an online questionnaire that is used to collect data relevant both for counseling and for program evaluation. (2) Admission takes place in the one-to-one chat of the program and is carried out by one of four qualified psychotherapists working for QTS. Usually, each user stays with the same counselor throughout the whole program. The admission chat takes ∼50 minutes, aims at clarifying the individual situation of the counseled person, and helps determine individual consumption goals and coping strategies. (3) After the admission chat, the online diary of QTS is activated. In this personal, password-protected area of the counseling program participants are required to write down all relevant aspects of their cannabis use over a period of 50 days. Once a week, all participants get detailed feedback on their entries by their counselor. The feedback relates to the consumption (current, in the course of the program), to the psychosocial situation of the participant, and to the counseling process as such. (4) The aims of the concluding chat are to reconsider to what extent the individual consumption goals were achieved, which strategies were suited to control cannabis consumption, and to determine whether further professional help is required. The program structure and counseling are based on the principles of self-regulation and self-control.²⁷ In the intervention, a solution-focused approach²⁸ is applied. So, all endeavors in the process of counseling focus on the aim to cease or reduce cannabis consumption. The program is free of charge. A secure connection to the program server via https-protocol is provided. The only personal data required for participation is the e-mail-address. Therefore, if an incognito e-mail-address is used, it is possible to utilize the program anonymously.

The study was conducted on the website of QTS (www.drugcom.de/?id=quittheshit) from December 2006 to August 2008. It was a two-arm-randomized controlled trial with one follow-up survey 3 months after entering the program. Subjects were randomized to either QTS or a waiting list.

To generalize the study results on the regular users of QTS, there were no predefined exclusion criteria. Everybody who expressed the wish to reduce or to cease cannabis use and wanted to take part in the study was included. To keep the threshold for a participation low and, thus, avoiding selection effects, personal data such as name, address, or phone number were not required, just like in everyday business of QTS.

Trial participants were recruited directly from the website of www.drugcom.de. At first, applicants were informed about the trial. In this context, they were advised about the possibility of not being accepted into the program immediately. To continue within the trial, participants had to agree to the study conditions (informed consent). All trial participants from the waiting list were given a binding commitment regarding program acceptance within 3 months. Moreover, they were informed about other help offers on www.drugcom.de.

After registering with the trial and filling out the baseline questionnaires (i.e., by entering the online-schedule for the admission chat), the participants were randomly allocated to the conditions by a software algorithm independent of the investigators and counselors. The follow-up survey was carried out 3 months after registration in an online-questionnaire. Recruitment was done by e-mail; thus, all QTS participants (regardless of the duration of the program usage) and all waiting list members were invited 3 months after trial registration. If participants did not react within 5 days, they were sent another invitation. This time, they were offered remuneration for their effort (shopping voucher worth 30 Euro).

At baseline, sociodemographic data, information on cannabis use, on use-related self-efficacy, and on psychological well-being was collected. Information on the frequency (days of consumption) and the quantity (in grams) of cannabis use during the last 30 days was gathered. The use-related self-efficacy was measured using a German adaptation of the Drug-Taking Confidence Questionnaire-8.²⁹ Cannabis addiction was determined in line with DSM-IV criteria published by the American Psychiatric Association.³ The Trait Scale of the State-Trait Anxiety Inventory³⁰ with a total of 20 items was used to measure the level of anxiety. Depressive symptoms was measured using the short version of the General Depression Scale (Allgemeine Depressions-Skala, ADS)³¹ consisting of 15 items. In addition, general life satisfaction was measured with the German adaptation of the Satisfaction with Life Scale (SWLS, 5 Items).³²

To examine the efficacy of the program, analyses of covariance were conducted. In each analysis, the baseline score of the corresponding outcome measure was included as covariate. In case of statistical significance, we calculated baseline-corrected Cohen's d as effect size.³³ An alpha level of 0.05 (two-tailed) was chosen. Per-Protocol- (PP) and intention-to-treat-analyses (ITT) were conducted. To deal with missing follow-up information, data of all randomized participants who did not complete the follow-up were imputed by Multiple Imputations (MI) and Last-Observation-Carried-Forward (LOCF) method. However, since MI usually provides more accurate parameter estimates of missing data than LOCF,^34,35 our ITT analyses concentrated on MI. To account for the sizeable drop-out rates, 100 imputations were conducted with Stata's “MI” command using the predictive mean matching method.

To determine differential loss to follow-up and to compare both study groups at baseline, logistic regressions were conducted (two-tailed α=0.05). All analyses were performed with Stata 11. After conducting a power analysis, a total sample size of n=200 was aimed for. We yielded to detect small to medium between group effects (d=0.35) with 80 percent power at an alpha-level of 5 percent.³³

1,292 persons declared that they were motivated to quit or at least significantly reduce cannabis use within the next weeks and, thus, registered with the study (Fig. 1). Of these, 429 persons were allocated to the waiting list and 863 to the treatment. To have comparable group sizes for the follow-up survey, two of three persons were allocated to the treatment condition. This unequal distribution was necessary to make up for the losses arising through persons who had been allocated to the treatment condition but who, in fact, did not take part in QTS (n=503). They had made use of the scheduler but then did not keep their appointment. Eventually, 360 persons participated in QTS. One hundred eighty-three of the QTS participants discontinued the intervention, that is, they quit the program before its duration of 50 days was reached. All 360 participants of QTS were invited for the follow-up.

Of the 429 persons allocated to the waiting list, 150 persons took part in the follow-up survey in due time, that is, within 15 days after receipt of the first invitation e-mail. In the treatment condition, 144 persons answered the questionnaire in the given period of time. 44 data records of the waiting list were excluded from analysis, as they either contained implausible information (n=1) or because the person completing the questionnaire claimed having made use of other professional help during the last 3 months (n=43), such as a drug advice center or a psychotherapist. Consequently, 106 cases were included in the analysis. In the treatment condition, 44 follow-up cases were excluded from the data analysis, mainly because they made use of additional professional help (n=25) or due to implausible information in the questionnaire (n=3). Further, 16 data records had to be rejected, because less than a minimum of 30 days had passed since the completion of QTS. This may happen if the registration and the admission chat are relatively far apart from each other due to time constraints. The analyzed sample in the treatment condition consists of 100 cases.

The mean age of the 206 follow-up participants (70.5 percent male) was 24.7 years (SD=6.8). On average, they consumed cannabis for 8.9 years (SD=6.2). The vast majority (92.0 percent) was cannabis dependent according to DSM-IV criteria. The educational level of the participants was relatively high,³⁶ with 58.7 percent attending grammar school (“Gymnasium”) or having reached A-Levels (“Abitur”). Participants of QTS used the intervention an average of 30.2 days (SD=19.8).

To detect differential loss to follow-up, bivariate logistic regressions were conducted with sociodemographic variables, cannabis use, and measures of psychological well-being as predictors and follow-up participation as outcome. Life satisfaction was the only significant predictor for follow-up participation (SWLS score: odds ratio: 1.02; 95 percent-confidence interval: 1.00–1.04; Wald Chi²(1)=4.07; p=0.044). Hence, participants with a higher score in the SWLS provided valid follow-up data slightly more often. Moreover, members of the treatment group who used QTS over the full 50 days were more likely to take part at follow-up than those who quit QTS prematurely (odds ratio: 3.80; 95 percent-confidence interval: 2.30–6.27; Wald Chi²(1)=27.06; p<0.001). The over-representation of those who completed the program as scheduled was balanced by applying a probability weight to the data during the following analyses. Since differential loss to follow-up had either been addressed by data weights or was just marginal, it was assumed that participants in the follow-up trial adequately represent all study participants.

No between-group differences in pretreatment scores were detected in any of the study measures.

In the PP analysis, which only included the data of those who left valid follow-up data, the analyses of covariances yielded significant effects in every outcome variable (Table 1). Users of QTS show a significantly stronger reduction in both consumption measures than members of the waiting list. The between-group effect sizes in use frequency and quantity can be considered as moderate to large. There are moderate effects on use-related self-efficacy. Further, QTS participants show a significantly stronger improvement of psychological well-being (Table 1). Thus, a lower level of anxiety and depression and a higher life satisfaction were observed in the treatment condition.

The ITT-analysis, which utilized MI, reveals statistically significant group differences in the use-related measures (frequency, quantity, and self-efficacy) and the level of anxiety (Table 2). However, the only effect that can be regarded clinically relevant was found for use frequency. In the LOCF-analyses, significant group differences use can be shown for the use-related measures too (frequency, quantity, and self-efficacy; p≤0.017; not shown in Table).