Many of the oSWT trials explored whether variables such as age, gender, education level or cultural background moderate treatment outcome. No evidence was found that the online treatment is more or less effective for specific demographic groups. Only two identifying predictive variables emerged (Lange et al. 2003a). Patients, who had never shared their traumatic events prior to treatment, were found to benefit more from oSWT than those who had. This finding confirms earlier evidence about the negative impact of not sharing traumatic events (Lange et al. 1999). Second, structured writing was found to be more effective with intentional traumatic events (i.e. those inflicted by neglect or intention), in comparison to non-intentional traumatic events (e.g. accidents or natural disasters).

In face-to-face CBT, patient ratings of the quality of the working alliance between the therapist and the patient have been found to correlate with treatment outcome (see, e.g. Martin et al. 2000; Norcross and Wampold 2011). Knaevelsrud and Maercker (2006) investigated the association of therapeutic alliance and outcome in online writing therapy, using the Working Alliance Inventory (WAI; Horvath and Greenberg 1989), a self-report measure of agreement on therapeutic tasks and therapeutic goals and on the degree of mutual trust and acceptance. Patients and therapists completed the WAI after phase I and immediately after treatment. Therapists’ alliance ratings did not change during treatment. Patients’ alliance ratings, however, did increase. Early alliance scores did not predict treatment outcome, although small to moderate inverse correlations were found between post-treatment patients’ alliance ratings and psychopathology assessments (i.e. IES-R, BSI; Knaevelsrud and Maercker 2007). In contrast, Wagner et al. (2012a) did find moderate to large correlations between mid-treatment WAI scores and reductions of core PTSD symptomatology of .28 (ns) < r < .49 (P < .01) in the context of the oSWT Ilajnafsy RCT. In sum, preliminary evidence currently suggests that online therapy does not preclude the development of an effective therapeutic working alliance. Yet, findings with regard to its relationship with treatment outcome are mixed.

Ruwaard (2009) explored the relation between treatment outcome and adherence to writing assignment instructions. He analysed a text corpus comprising 4255 essays written by the 478 traumatised patients that were included in the routine practice study (Ruwaard et al. 2012). He used Pennebaker’s Linguistic Inquiry and Word Count (LIWC; Pennebaker et al. 2007), a programme for automated content analysis that calculates the occurrence of different categories of words. LIWC scores clearly mirrored assignment instructions. For example, frequencies of LIWC’s ‘I’ word category strongly decreased from phase I to phase II, while ‘You’ words inversely increased, in line with the therapeutic instruction to change perspective in phase II. However, these LIWC scores did not convincingly predict treatment outcome as measured by the self-report questionnaires. After Bonferroni corrections for multiple statistical testing, no statistically significant relationships between LIWC scores and treatment outcome remained.

Lemmen and Maas (2001, see Lange et al. 2003) applied conventional content analysis to study the relation between the development of cognitive coping during treatment and treatment outcome. They created contrast groups out of 101 patients, who had completed writing treatment for PTSD: ten patients who had benefitted most and ten patients who had benefitted least. Independent judges, blind to the groups, rated essays 1, 3, 8 and 10 (presented in random order) on several aspects of cognitive coping: insight in the process of overcoming the traumatic events (e.g. showing reflection on dysfunctional automatic thoughts), functional coping by expressing behavioural adaptation (e.g. decrease in avoidance behaviour, showing more assertiveness; Donnelly and Murray 1991) and orientation on the future (e.g. planning positive things, using words such as ‘I will’). In both groups, the increase in coping was significant and large (F _3,54 = 40.1, P < .001). As expected, essays did not show an increase in coping during phase I. However, essays written in phase II (cognitive reappraisal) and phase III (closure/social sharing) reflected large increases in coping. Increases in functional coping were larger in the most improved group in comparison with the least improved group, although these differences were not significant (F _1,19 = 2.2, P < .15). Since the group, that had profited the least, also showed large benefits in coping, variance in outcome was very small. This may explain the non-significant difference between the two groups.

Van Emmerik et al. (2008) compared face-to-face structured writing therapy (SWT; n = 44) to regular face-to-face CBT (n = 41) and WLC (n = 40) in an RCT that included 125 adults with a confirmed diagnosis of acute stress disorder or PTSD. Structured writing therapy closely followed the structure of the original online structured writing protocol. Writing assignments were provided as homework assignments and were discussed in five to ten face-to-face feedback sessions. There were no statistically significant differences between the two active treatment conditions on any of the outcome measures (CBT vs. SWT, .03 < d < .4). Both active treatments, however, were superior to WLC (.3 < d < .7).

Van der Oord et al. (2010) adapted oSWT into a standardised computer-aided treatment for traumatised children and adolescents (‘Write Junior’). In this ‘blended’ treatment, the young patient completes the writing assignments on a computer during face-to-face sessions, with the help of the therapist, to construct a coherent narrative of the traumatic event, in which elements of self-confrontation, reappraisal and social sharing are integrated. For very young children, drawing replaces computer-based writing. In a small uncontrolled pilot trial of this protocol variant (n = 23; age of participants: M = 12 years), strong, significant and stable reductions in PTSD symptomatology were found.

Seitz et al. (2014) explored the application of a variant of oSWT in the treatment of young adult survivors of paediatric cancer (age, 20–26 years) with elevated symptoms of PTSS and/or anxiety, in a small (N = 28) uncontrolled study of what they call Onco-STEP. As the original oSWT protocol, Onco-STEP consists of ten therapist-assisted online structured writing exercises. However, the phases of treatment are structured differently. In the first phase – ‘Looking back’ (five assignments) – patients address the traumatic cancer-related experience through imaginary exposure and cognitive reappraisal. In the second phase, ‘Looking forward’ (five assignments), patients address dysfunctional fears of disease progression. In a completer analysis (n = 20), participants reported moderate and significant reductions in symptoms of PTSD, anxiety, fear of disease progression and depression (.5 < d < 1.0). These results suggest that structured online writing may also be suitable to provide online psychological support to cancer survivors.

Structured writing exercises have been applied in several variants in other Internet-based CBT programmes for PTSD. Hirai and colleagues examined unguided online writing. In the first RCT (N = 27; Hirai and Clum 2005), they assessed the effects of a self-help ICBT programme that combined relaxation training, cognitive restructuring and exposure through writing. Moderate to large between-group effect sizes were found in favour of ICBT in comparison to a waiting-list comparison group, although between-group effects were significant only with regard to depression, state anxiety, avoidance symptoms and frequency of intrusions (.8 < d < 1.3). In a later RCT (N = 104; Hirai et al. 2012), they found indications that online writing induces better effects when patients are instructed to write not only about facts of the traumatic event but also about related emotions (as in oSWT). Litz et al. (2007) compared therapist-guided online CBT to online supportive counselling in an RCT (N = 45). Both treatment programmes included repeated writing. In the ICBT group, writing focused on identified traumatic events. In the supportive counselling group, writing focused on daily concerns. With ICBT, the researchers found a sharper decline of PTSD symptoms over time and higher rates of end-state functioning at 6-month follow-up. In contrast, Possemato et al. (2011) did not find significant differences between trauma-focused vs. non-trauma-focused online writing in a small RTC (N = 31), although this could have resulted from the low power of the study and the fact that patients completed only three writing sessions. Klein et al. (2010) further added to the evidence by demonstrating promising effects of a 10-week therapist-guided ICBT programme (including online writing) in a small uncontrolled trial (N = 22) that included patients with a primary clinical diagnosis of PTSD. Pre-post-follow-up assessments revealed significant reductions of PTSD severity and related PTSD symptomatology (although changes in general psychopathology ratings were non-significant). This study also suggested that therapist input might be limited in ICBT, since the reported average duration of therapist guidance during treatment was only 194 min (which the authors contrasted to a standard of 12 h in face-to-face CBT for PTSD; Harvey et al. 2003). More recently, these results were replicated and extended through waiting-list controlled trials by Spence et al. (2011) and Ivarsson et al. (2014). In both studies, large and significant effects of therapist-guided ICBT were found with regard to core symptoms of PTSD and symptoms of general psychopathology, in patient samples with confirmed diagnoses of PTSD.