Table gives a demographic overview of the sample, separately for each treatment condition. There were no systematic group differences in any of the socio-demographic variables. Patients of the counselling group reported no use of medication (herbal medication included) during the treatment or follow-up period and patients of the medication group (routine medical treatment) did not take any non-prescribed medication. For a detailed description of the prescribed medication see Table .
In the counselling group, 3 patients (8.8%), one being male, dropped out of treatment. In the medication group (routine medical treatment), 2 patients (6.3%), both being male, did not complete treatment (see flow chart in Figure ). Men dropouts of both treatment conditions reported not being able to afford the time to continuously visit the distantly located health care centre.
Flowchart of the study protocol.
We found a significant interaction of treatment × time in the HSCL depression score (F(1,59) = 175.21, p < .001, ηp2 = .75; which also produced a main effect (F(1,59) = 142.60, p < .001, ηp2 = .71). As illustrated in Figure , pre-treatment depression scores were the same for the two groups (counselling M = 41.65, SD = 6.03 and medication group M = 43.00, SD = 6.53; two sample t-test, t(60) = -.84, p = .40, d = .22). At 3-month follow-up, the psychosocial counselling group showed significantly lower HSCL depression scores (M = 20.26, SD = 1.95) than the medication group (M = 44.10, SD = 5.64) (Mann-Whitney U test, U = .00, z = -6.73, p < .001, r = 0.86). While we found a large treatment effect size for the psychosocial treatment group in the reduction of the HSCL depression score (M = -21.39, SD = 6.54, one sample t-test, t(30) = -18.21 p < .001, d = 3.27), the change in the HSCL depression score in the medication group indicated that there was no improvement (M = 1.10, SD = 6.73, one sample t-test, t(29) = .90, p = .38, d = .163). A last observation carried forward analyses considering participants that dropped out revealed a similarly large significant treatment effect (Mann-Whitney U test, U = 16.50, z = -6.79, p < .001, r = 0.84).
Change in HSCL-Depression/-Anxiety scores. a. Change in HSCL-Depression scores. b. Change in HSCL-Anxiety scores.
Similar results were obtained for the anxiety scores (Figure ). In a repeated measures analysis of variance (ANOVA) with the HSCL anxiety score at pre-treatment and follow-up as the within-subject factor and psychosocial counselling versus medication as between-subject factor was calculated. Again, a significant interaction indicated that the HSCL anxiety score decreased in the counselling but not the medication group (F(1,59) = 172.46, p < .001, ηp2 = .75; main effect F(1,59) = 198.89, p < .001, ηp2 = .77). Both groups had nearly identical values in their HSCL anxiety scores at pre-treatment (psychosocial counselling group: M = 29.52, SD = 4.63, medication group: M = 30.63, SD = 4.22, two sample t-test, t(60) = -.983, p = .329, d = .25). At the time of post-treatment, the counselling group showed lower HSCL anxiety scores (M = 12.68, SD = 1.33) than the medication group (M = 30.03, SD = 5.13) (Mann-Whitney U test, U = 1.00, z = -6.74, p < .001, r = 0.86). A last observation carried forward analyses revealed the treatment effect were about the same when drop outs are considered in the calculation (Mann-Whitney U test, U = 19.00, z = -6.78, p < .001, r = 0.83). Again, the reduction of the HSCL anxiety scores revealed a large treatment effect size in the psychosocial counselling group (M = -16.84, SD = 4.87, one sample t-test, t(30) = -19.24, p < .001, d = 3.46), while the change in the medication group was negligible (M = -.60, SD = 4.78, one sample t-test, t(29) = -.69, p = .497, d = .125). Thus, only the psychosocial counselling significantly improved the depression and anxiety symptoms.
These findings are validated by further assessments through the M.I.N.I. and the Screening for Depression. Pre-treatment diagnoses of current major depression assessed through the M.I.N.I. did not differ between both groups (counselling group N = 27 (87.1%), and medication group N = 27 (90%); χ2 (1) = .17, p = 1.00). The percentage of counselling treatment patients meeting M.I.N.I. criteria for a diagnosis of a current major depression dropped to 0%, whereas 28 (93,3%) patients of the medical treatment met M.I.N.I. criteria for such a diagnosis at follow-up. The counselling and medication group significantly differed in the status of diagnosis for current major depression at follow-up (χ2 (1) = 56.12, p < .001).
The results of the Screening for Depression also show that in the counselling group the depression score significantly changed between pre-test and follow-up (M = -10.97, SD = 2.87, one sample t-test, t(30) = -46.97 p < .001, d = 3.82). At the same time, the medication group showed an increase in symptom severity (M = 1.50, SD = 2.87, one sample t-test, t(29) = 2.81, p < .01, d = .57). Additionally the Screening for Depression shows high correlations with the MINI (r = .424**) and the depression section of the HSCL 25 (r = .682**).
Besides depression and anxiety symptoms, we assessed the psychosocial stressors reported by the patients. The average number of reported current psychosocial stressors was 3.51 (SD = 1.41) for the whole sample. There was no significant difference between the groups before the treatment (psychosocial counselling group: M = 3.29, SD = 1.37, medication group: M = 3.20, SD = .96, two sample t-test, t(60) = -.30 p = .768, d = .08).
The most frequent psychosocial stressor types were, family conflicts (n = 47; 77%) and (inter)personal problems and difficulties such as issues of honour and shame (n = 41; 67.2%). In addition, ongoing domestic violence appeared to be not unusual among interviewed patients (n = 16; 26.2%). After the treatment, the patients in the counselling group reported fewer psychosocial stressors (M = .74, SD = .68) than the patients in the medication group (M = 3.57, SD = 1.01) (Mann-Whitney U test, U = 8.00, z = -6.71, p < .001, r = .86). Apart from poverty (35.5%), all reported stressors dropped to under 5% in the counselling group. Figure provides a more detailed look at the specific types of psychosocial stressors and specific changes within treatment time for both groups.
Counselling resulted in a large effect size for the reduction in the psychosocial stressors (M = -2.55, SD = 1.18, one sample t-test, t(30) = -12.04, p < .001, d = 1.91), whereas the patients of the medication group felt that their psychosocial stressors occurred more frequently at follow-up (M = .37, SD = .61, one sample t-test, t(29) = 3.23, p = .003, d = .61).
The effect of psychosocial stressors on the symptom reduction
We found a high correlation between the reduction of psychosocial stressors and the symptoms for both HSCL scores, depression (rp = .81, p < .001) and anxiety (rp = .82, p < .001). To analyse whether there was a treatment effect on the HSCL depression change in the counselling group besides the one reduction in psychosocial stressors, we performed a mediation analyses for the mediator psychosocial stressors. For the depression symptoms, the change in the HSCL depression score between pre-treatment and follow-up (HSCL depression change) was regressed on the change in psychosocial stressors (psychosocial stressors change) as well as on the treatment that was dummy coded with "0" for the medication group and "1" for the psychosocial counselling group in a first step. In a second step, the psychosocial stressor change was regressed on treatment.
Figure shows that the reduction in the number and frequency of psychosocial stressors contributes to the reduction in the HSCL depression score (βpsychosocial stressor change
= .26, p
= .031) as indicated by the high correlation between these two variables. Still, as the treatment also accounts for the reduction in psychosocial stressors (βtreatment
= .-84, p
< .001), the treatment outcome in the HSCL depression score changes is therefore mediated via this indirect treatment effect. Moreover, there is also a direct treatment effect beyond the reduction of psychosocial stressors accounting for changes in the HSCL depression score (βtreatment
= .-65 p
< .001). These two variables are responsible for 76% of the variance in the HSCL depression score changes (Radj2
= .76, F
(2,58) = 95.87, p
< .001, f2
= 3.17). The sample size for the regression analysis was sufficient as indicated by the calculated power ((1 - β)
= 1.00) using G × Power 3 [26
]. Moreover, a further analysis of the residuals on collinearity, normal distribution of the residuals, homoscedasticity and influence on outliers indicated that the proposed model fulfils all necessary quality criteria. Additionally, the interaction term treatment × psychosocial stressor change did not explain more variance significantly F
(1,57) = .31, p
= .579) indicating that the psychosocial stressor change mediates but not moderates the treatment effect. Taken together, the psychosocial counselling had a large effect on the HSCL depression score changes.
Figure 4 The effect of psychosocial stressors on the treatment outcome regarding depression and anxiety. a. The effect of psychosocial stressors on the treatment outcome regarding depression. b. The effect of psychosocial stressors on the treatment outcome regarding (more ...)
The same analysis was calculated for changes in the HSCL anxiety score (Figure ). The sample size was sufficiently large for the regression analyses ((1 - β) = 1.00).
The reduction in the psychosocial stressors contributed to the reduction in the HSCL anxiety score (βpsychosocial stressor change = .26, p = .031), likewise to the HSCL depression score. Also, the treatment accounted for the reduction in psychosocial stressors (βtreatment = .-84, p < .001) and consequently had an indirect effect on the treatment outcome via this path. Moreover, also a direct treatment effect accounted for changes in the HSCL depression score (βtreatment = -.60, p < .001). The regression model that included the two variables, treatment and psychosocial stressor change, had a large effect on the variance of the HSCL anxiety score (Radj2 = .77, F(2,58) = 98.51, p < .001, f2 = 3.35). The power in this model was sufficient ((1 - β = 1.00). As for the preceding model on HSCL depression change, an analysis of the residuals revealed that all necessary quality criteria were met. Moreover, there was no moderation effect of the psychosocial stressors.
Besides psychosocial stressors, we assessed coping mechanisms of the whole sample, with 56 (91.8%) patients being unable (scoring 0 and 1 on a 4-point likert-scale) to recognize a relationship between own symptoms and specific psychosocial stressors before treatment. At the same time, 55 (90.2%) patients were not able to manage or solve current conflicts (scoring 0 and 1 on a 4-point likert-scale). Again, at follow-up there was a significant between-group difference (t(59) = -28.58; p < .01, with counselling patients showing improved coping strategies in all categories at follow-up (M = 2.39; SD = .34; t(30) = 28.33; p < 0.01). Such a change did not occur in the medication group (routine medical treatment) (M = .33; SD = .21; t(29) = -8.43; p < .01).