Five-year outcomes in a randomised controlled trial of prolonged exposure therapy and supportive counselling for post-traumatic stress disorder in adolescents: a task-shifted intervention Authors: Jaco Rossouw, Elna Yadin, Debra Alexander, Soraya Seedat Published: 2026 ABSTRACT ABSTRACT BackgroundCognitive–behavioural therapies with a trauma focus are effective in reducing posttraumatic stress disorder and other psychological distress in adolescents. Long-term follow-up data on adolescents treated for PTSD remain scarce, with few studies extending beyond 12 months after treatment completion.ObjectiveTo evaluate the maintenance of treatment gains in a comparative study of effectiveness of PE-A and SC up to 60 months post-treatment.MethodSixty-three adolescents diagnosed with PTSD were randomly assigned to either treatment, provided by newly trained and supervised non-specialist health workers. The primary outcome measure was PTSD symptom severity, as independently assessed on the Child PTSD Symptom Scale (CPSS). We report on the 60-month post-treatment follow-up, building on post-treatment, 3-month, 6-month, 12-month and 24-month post-treatment data that have been published previously.ResultsParticipants in both treatment groups maintained a significant reduction in PTSD symptoms up to 60-months post-treatment (F (7, 343) = 2.86, p < .01). Participants receiving prolonged exposure experienced greater improvement on the CPSS at all follow-up assessment timepoints, except for the 60-month FU (p = .28; g = 0.33).ConclusionAdolescents with PTSD continued to maintain treatment gains up to 60-months post-treatment. These data, along with findings from the original RCT, indicate that a brief treatment protocol (averaging 9 sessions of PE-A or SC) in a LMIC, task-shifted to be delivered by nurses without prior psychotherapy experience, led to lasting improvements in PTSD and comorbid symptoms for up to five years. The sustained benefits and improved functioning over the first few years post-treatment support expanding both treatments, especially PE-A, in community settings. Background Trauma exposure in South African adolescents is highly prevalent, with 50-90% of adolescents reported to have experienced at least one traumatic event (Burton et al., 2015; Burton & Leoschut, 2013; Seedat et al., 2004; Suliman et al., 2005). Most adolescents are exposed to violence in their own communities (Kaminer et al., 2013; Otwombe et al., 2015) and as many as 90% of South African youth experience poly-victimisation, with exposure to multiple forms of violence (Collings et al., 2014; Herrero-Romero et al., 2021). In a national survey of South African adolescents (Ward et al., 2018), prevalence rates by trauma type were 12% for sexual abuse, 18% for physical abuse, and 25% for exposure to family violence. In a report by UNICEF, more than 50,000 children in South Africa were victims of violent crimes between 2011 and 2012 (Viviers, 2013). Alarmingly, in a more recent report, a staggering 99% of children in some areas experienced or witnessed physical or sexual violence in their home, school, or community (Richter et al., 2018). Posttraumatic Stress Disorder (PTSD) can develop secondary to trauma exposure and, if left untreated, can become a serious chronic condition. The rate of PTSD among South African adolescents has been reported to range between 8% and 22% (Burton & Leoschut, 2013; Seedat et al., 2004; Suliman et al., 2005). This high rate of PTSD in this vulnerable population of adolescents underscores the need for effective, palatable, and sustainable preventive and therapeutic interventions. Task-shifting has been proposed as one strategy to reduce the mental health treatment gap in low- and middle-income countries (LMICs), where specialist mental health resources and access to evidence-based psychological interventions remain sparse. In our previous work, a feasibility study showed that Prolonged Exposure for Adolescents (PE-A) and Supportive Counselling (SC) could be successfully delivered by newly trained and supervised non-specialist health workers in a community-based, school setting, with both interventions perceived as acceptable and feasible (Rossouw et al., 2016). A subsequent randomised controlled trial (RCT) demonstrated that nurses previously naïve to providing psychotherapy could be trained and supervised to effectively deliver both PE-A and SC, with PE-A producing significantly greater reductions in PTSD symptoms than SC at post-treatment and at 3- and 6-month follow-up (Rossouw et al., 2018). Qualitative studies nested within this research further indicated that school-based, task-shifted PTSD interventions were generally acceptable to adolescents, the nurse providers, and to other stakeholders, while also highlighting implementation considerations such as stigma, confidentiality, access, supervision, and fit with school routines (Van der Water et al., 2017; Van der Water, Rossouw, van der Watt et al., 2018; Van der Water, Rossouw, Yadin et al., 2018). Together, these findings support task-shifting as a potentially feasible and scalable approach to expanding access to evidence-based, trauma-focused psychological treatments for adolescents in low-resource settings. Cognitive behavioural therapies with a trauma focus (TF-CBT) have been found to be efficacious treatments for young people with posttraumatic stress symptoms, as confirmed in a recent meta-analysis on individual participant data (de Haan et al., 2024). These therapies were found to be superior to both active and passive control conditions in reducing PTSD symptom severity. However, in treatment studies of young people with PTSD, despite the importance of determining the durability of effects, studies rarely report longer-term follow-up (FU) assessments beyond 12 months (Gutermann et al., 2016; Rith-Najarian et al., 2019). One exception was a study that reported long-term FU of a randomised controlled trial of task-shifted prolonged exposure therapy (PE-A) and supportive counselling (SC) for PTSD in adolescents (Rossouw et al., 2022). This study showed that adolescents who had been diagnosed with PTSD and who received PE-A or SC provided by non-specialist health care workers in a community setting maintained a pattern of treatment gains when assessed 12 and 24 months after treatment completion (Rossouw et al., 2022). The current paper extends that study (Rossouw et al., 2022) by reporting findings from the 60-month FU assessment in the same cohort of adolescents who had received either PE-A or SC for PTSD. In order to depict symptom trajectories over time, 60-month follow-up data are reported here in the context of all assessment points already reported previously (Rossouw et al., 2018, 2022). The main objective of this study was to evaluate the long-term durability of treatment gains in the two groups who had received PE-A or SC for PTSD as adolescents. We hypothesised that both treatment groups would maintain treatment gains over time and that PE-A would retain its superiority over SC in reducing both interviewer-assessed PTSD severity and measured secondary outcomes across follow-up assessments. In addition, as it has been shown that, based on patient self-report, PE-A and SC are both effective in reducing many adolescent problems that co-occur with PTSD (Zandberg et al., 2016), we hypothesised that positive changes in anxiety, mood regulation, self-esteem, resilience, perceived social support and anger would also be found and maintained long-term. Methods A description of the methods, previously reported in detail (Rossouw et al., 2018, 2022), is briefly presented below. Participants Adolescents between the ages of 13–18 years with chronic PTSD (> 3 months) were recruited from schools around Cape Town, South Africa. Adolescents with a primary diagnosis of PTSD with or without a comorbid mood, anxiety, attention-deficit disorder, or substance use disorder were included. Adolescents with a primary substance disorder, psychotic disorder or conduct disorder were excluded from the trial. Procedure Ethics approval for this study was obtained from the Health Research Ethics Committee of Stellenbosch University (N12/06/031). The Western Cape Education Department (WCED) approved the study, conditional on not disclosing the names of the 11 low-income community schools. Findings from our pilot study (Rossouw et al., 2016) indicated that recruitment of study participants required raising awareness of PTSD as a diagnosis and of the benefits of psychological treatment. Recruiters visited community schools and presented this information either at the class level, to all classes in a grade, or to the entire school. Adolescents who were interested in participating in the trial contacted the coordinator at the school and completed a PTSD screening instrument. Adolescents who described an interpersonal trauma that they experienced or witnessed at least 3 months earlier were scheduled for a baseline assessment, accompanied by a parent or a guardian. A full description of the assessment instruments used in the studies can be found in the section on Assessment and Outcome Tools. Those adolescents who scored above the cutoff score for inclusion, with or without comorbid mood disorders, anxiety disorders, substance use and attention-deficit hyperactivity disorder were eligible for inclusion, provided that PTSD was determined to be the primary disorder requiring treatment. Adolescents with conduct disorders, primary substance use disorders or psychotic disorders were excluded. All assessments were conducted by two experienced independent evaluators (IEs), blinded to treatment condition. The baseline assessment comprised a structured clinical interview and the completion of self-report measures (2–3 h). Participants who met criteria for PTSD and who met the inclusion and exclusion criteria described earlier were included in the study. Follow-up assessments included a structured assessment of the presence of a PTSD diagnosis and completion of self-report measures (1.5 h). A sample size of 64 participants was calculated to provide 80% power for a two-tailed alpha level of 0.1 to detect a medium effect size. Eligible participants were randomised to one of two treatment arms (either PE-A or SC) using a parallel, permuted block procedure with six randomisations per block (1:1 ratio). Following is a description of the two treatments offered in the original RCT. PE-A treatment The PE-A intervention (Foa et al., 2009) is a modular treatment tailored for youth. Seven to fourteen weekly sessions, each 60 minutes in duration, comprised (1) an explanation of the rationale for treatment, information gathering, identifying the index trauma, and conducting a breathing retraining exercise; (2) discussion of common reactions to trauma; discussion of the rationale for in vivo exposure (facing trauma reminders in real life), construction of an in vivo hierarchy, and assignment of in vivo homework; (3) presentation of the rationale for imaginal exposure (remembering and recounting the traumatic memory), conducting imaginal exposure, and processing the experience and the meaning of events that were revisited; (4) focusing on the worst moments of the trauma; discussion of the generalisation of skills learned in treatment and a focus on relapse prevention; and (5) ending with a ‘final project’, detailing the trauma, and gains made in treatment. Treatment completers were defined as having completed a minimum of seven sessions which would have ensured that participants in the PE-A treatment group received all the main components of the evidence-based treatment across the sessions (Rossouw et al., 2018). SC treatment Supportive Counselling (SC), the non-trauma-focused, comparator intervention (Cohen & Mannarino, 1996), was delivered in an earlier adolescent treatment trial (Foa et al., 2013). The manualised SC time-matched intervention consisted of 7–14 weekly, 60-minute sessions. SC is a client-centered treatment, focussed on establishing a trusting, empowering, and validating relationship. After an initial orientation to the treatment during session 1, participants set the present-centered agenda for each session. Participants were allowed to decide whether and when to discuss the trauma. SC treatment providers noted whether a discussion of the trauma was initiated and the time devoted to this. No participants, except for one prompted by a counsellor, described their trauma during SC sessions. Treatment entailed providing active listening, empathy, encouraging expression of emotion, and non-directive problem solving. Assessment and outcome tools A demographic questionnaire, capturing socio-demographic information on the participants and their caregivers, and the full Mini International Neuropsychiatric Interview for Children and Adolescents (MINI-KID; Sheehan et al., 2010) were administered only at the baseline assessment. The MINI-KID is a semi-structured interview administered by an independent clinical psychologist evaluator with extensive experience in diagnostic assessment in RCT trials for PTSD, covering current and lifetime disorders. It has excellent test-retest reliability (0.64–1.00). The full MINI-KID was used to determine the presence of a diagnosis of PTSD and other diagnoses at baseline. At all post-interventions independent evaluator assessments, the PTSD component of the MINI-KID was used to confirm the presence or absence of a PTSD diagnosis. Primary outcome measure The Child PTSD Symptom Scale–Interview (CPSS-I; Foa et al., 2001; Gillihan et al., 2013) was the primary outcome measure of PTSD symptom severity and was administered as an interview by an IE at all assessment points. A cutoff total score of 11 or higher on the CPSS-I yielded 95% sensitivity and 96% specificity of diagnosis. Secondary outcome measures The Child PTSD Symptom Scale – Self Report (CPSS-SR; Foa et al., 2001; Gillihan et al., 2013) and the Beck Depression Inventory (BDI; Beck et al., 1961, 1988) were used to track changes in PTSD- and mood symptom severity across treatment sessions. These measures were also used at all FU assessments. The BDI was previously used with adolescents in a South African context (Ward et al., 2003). The Children’s Global Assessment Scale (CGAS; Shaffer et al., 1983) was administered by the IE at pre-treatment as well as at post-treatment and at all FU assessments as a summary measure of functional impairment. Additional secondary outcome measures Additional instruments were administered throughout to assess for potential comorbid conditions typically seen in adolescents. These secondary outcomes have not been reported in previously published papers and are, therefore, described in greater detail, below: The Multidimensional Anxiety Scale for Children (MASC; March et al., 1997) is a 39-item scale that assesses a spectrum of common anxiety symptoms in youth. Items are distributed across 4 major factors: physical symptoms, social anxiety, harm avoidance, and separation anxiety; the ratings are done on a 4-point Likert scale, with higher scores denoting higher anxiety. The scale has been shown to be stable across gender and age and has demonstrated excellent internal consistency with Cronbach's alpha scores generally above .70, test-retest reliability ranging from .62 to .93 over various time intervals, and established convergent validity (March et al., 1997). A study showed that the same 4-factor structure applies to adolescents in the Cape Town metropole when compared to North American and Icelandic samples (Fincham et al., 2008). The Negative Mood Regulation questionnaire (NMR; Catanzaro & Mearns, 1990) is a 30-item self-report measure of beliefs that individuals maintain that can adversely impact their ability to feel better. Higher scores indicate better expectancies around the ability to control negative mood. Internal consistency was excellent, ranging between .87 and .92, test-retest reliability was acceptable, ranging between .73 and .74 for a 4-week interval and .67 and .78 for a 6- to 8-week interval. The Rosenberg Self-Esteem Scale (RSES; Rosenberg, 1965, 1979) is a self-rating 10-item scale to assess self-esteem, with higher score denoting better self-esteem. The RSES demonstrated excellent reliability, with internal consistency of .92, stability, with test-retest reliability over a period of 2 weeks of .85 to .88, as well as concurrent, predictive and construct validity. The scale has been evaluated across multiple cultural contexts (Baranik et al., 2008), including Western, Eastern and African contexts. The Connor-Davidson Resilience Scale (CD-RISC; Connor & Davidson, 2003) is a 25-item self-report measure to assess resilience and demonstrated good internal consistency, with a Cronbach's alpha of .87–.89, test-retest reliability of .87, and good convergent/divergent validity. The scale has been validated across various populations and cultures, indicating its effectiveness in different contexts (Wojujutari et al., 2024). The Multidimensional Scale of Perceived Social Support (MSPSS; Zimet et al., 1988) is a 12-item self-report inventory, using a 7-point Likert scale, that measures the perceived adequacy of social support from 3 sources: family, friends and significant others. The psychometric qualities of the MSPSS were evaluated within a South African context with the internal reliability of the scale and subscales found to be high, with Cronbach alpha coefficients > 0.85 and good test-retest reliability (Bruwer et al., 2008). The Spielberger State-Trait Anger Expression Inventory-2 (STAXI-2; Spielberger et al., 1970, 1999) is a 57-item self-report measure, using a 4-point Likert scale. It has a State-Anger subscale measuring the intensity of angry feelings and the extent to which a person feels like expressing anger at a particular time, and a Trait-Anger scale that measures how often angry feelings are experienced over time. Anger control is measured by presenting descriptions of responses and participants rate how often they behave in the manner described when angry. High scores on the STAXI-2 scales indicate anger experience and expression that interferes with normal functioning. Norms are available for adolescents 16–19 years of age. Internal consistency is good, with a Cronbach alpha of .88, good test-retest reliability, ranging from 0.64 to 0.82 for a two-week interval. In the original study (Rossouw et al., 2018), an Expectancy of Therapeutic Outcome for Adolescents scale (ETO-A; Foa et al., 1991) was completed after the rationale for the treatment was explained. It is a brief adaptation of the 4-item adult ETO (Foa et al., 1991), derived from Borkovec and Nau's (1972) credibility items, assessing the adolescent’s belief about the positive outcome of the treatment. The ETO-A asked the adolescents to rate their expectations on a 5-point Likert scale ranging from ‘I expect this treatment to help me a lot’ (5) to ‘I don’t expect this treatment to make any difference in my condition’. In addition, a measure of perceived treatment helpfulness assessing satisfaction with treatment outcome (STO) was completed by the participant at all FUs and included statements on a 5-point Likert scale, ranging from ‘I am extremely satisfied with this treatment, it helped me a great deal’ (5) to ‘I am not at all satisfied with this treatment, it did not help my condition’ (1). No factor-analysis, validity, or reliability studies of the STO have been published. A measure of therapeutic alliance (WIA-CA) was completed by the adolescent at every FU assessment and included statements such as ‘[Therapist name] and I agree about the steps to be taken to improve my situation’ and was rated on a 7-point Likert scale ranging from ‘Always’ (7) to ‘Never’ (1). The Working Alliance Inventory for Children and Adolescents (WAI-CA; Figueiredo et al., 2019) is a 36-item self-report measure of therapeutic alliance in children and adolescents, covering goals, tasks, and bond. It has demonstrated acceptable to high internal consistency (Cronbach’s alpha = from .71 to .89), and external validity in youth psychotherapy samples. Training and supervision Treatment providers were nurses who were non-specialist health workers (NSHWs), previously naïve to providing psychological treatments. They were at the start of a one-year diploma course in psychiatric nursing at Stellenbosch University. The training of nurses’ in both treatments, as described in detail in the Procedure section, consisted of a 3-day PE-A workshop, and a 1-day SC training plus 2 days of roleplaying, where they received training in the theoretical basis and practical aspects of each treatment. To minimise potential provider effects and to prevent individual counsellors from being confounded with treatment condition, all nurses were trained in both interventions and delivered both treatments. A permuted block randomisation procedure ensured balanced allocation of PE-A and SC cases across providers. Group supervision was conducted weekly throughout the active treatment phase of the study. Individual supervision was available to address case-specific management issues and to enhance adherence. Data analysis Linear mixed models were used because they account for within-participant correlation across repeated assessments and yield unbiased estimates under intent-to-treat with data missing at random (Gueorguieva & Krystal, 2004; Laird & Ware, 1982) and were fitted using the lmer package in R (Bates et al., 2015). Post-hoc contrasts were conducted using Fisher's LSD, which maintains familywise error control in limited comparison sets while preserving power in small samples (Fisher, 1935; Levin et al., 1994). Effect sizes were also reported. To evaluate differences between the two groups over time in the proportion of participants meeting a PTSD diagnosis, we used generalised estimating equations (GEE) with binomial as the underlying distribution. GEE analyses were done using the R package ‘geepack’ version 1.2-1. Because the GEE is a quasi-likelihood method, likelihood-ratio tests are not defined. Therefore, fixed effects were tested with Wald statistics (Wald, 1939) based on the empirical (sandwich) covariance matrix. The Group × FU interaction was assessed with a multiparameter Wald test of the six interaction coefficients. Baseline differences between treatment groups were examined for demographic and clinical variables using independent-samples t-tests for continuous variables and chi-square tests for categorical variables. No significant differences in any baseline scores between the treatment groups were found, and, consequently, no covariates were included in the analyses of primary and secondary outcomes. Results Of the 161 adolescents who volunteered to participate in the study, 133 completed baseline assessments (February 2014 to June 2016). Sixty-three adolescents were eligible for the study and were randomised to receive either PE-A (n = 31) or SC (n = 32) (Figure 1). Figure 1.Consort flow diagram.A flowchart diagram showing screening of 161 adolescents, exclusion steps, randomization of 63, and follow up in PE-A and SC groups.The figure shows a flowchart diagram tracking adolescents through a randomized study comparing Prolonged Exposure for Adolescents, abbreviated as PE minus A, with Supportive Counseling, abbreviated as SC. At the top, a rectangle states that 161 individuals are screened for presence of trauma and more than 3 months since trauma. A right arrow leads to a rectangle listing 28 excluded, with counts for did not meet inclusion criteria, no trauma event, grief response, and not contactable. A downward arrow from the initial box leads to a rectangle noting 133 scheduled for baseline assessment to assess eligibility and complete consent and assent. A right arrow from this box leads to a rectangle listing 70 excluded, including did not meet inclusion criteria, required social service intervention, met exclusion criteria, could not get parental consent, and required referrals to other services. A downward arrow from the baseline box leads to a rectangle noting 63 randomized. The diagram then splits into 2 arms. The left arm shows 31 randomized to receive Prolonged Exposure for Adolescents, with 25 completing treatments more than 7 sessions, followed by boxes listing numbers completing mid assessment, post assessment, 3 month, 6 month, 12 month, 24 month, and 60 month assessments, and ending with 31 included in analysis. The right arm shows 32 randomized to receive Supportive Counseling, with 27 completing treatments more than 7 sessions, the same series of assessment completion counts, and 32 included in analysis. Consort flow diagram. Demographic data for the sample is provided in Table 1. The index trauma types experienced included sexual assault (49%), physical assault (19%) or witnessing an assault (31%). There were no differences in adolescents’ expectancy scores of PE-A and SC measured by the ETO total scores (p = .80). The dropout rate during treatment was acceptable (< 20%) in both treatment arms (p = .71). Both treatment groups completed a mean number of 9 sessions, and the mean session length was 47.78 min (PE-A = 46.55; SC = 48.87; t = −1.73; p = .042). Table 1.Demographic variables of the intent-to-treat sample (N = 63).VariableTotal (N = 63)PE-A (n = 31)SC (n = 32)pAge, M (95% CI)15.35 (14.98, 15.72)15.19 (14.66, 15.72)15.50 (14.97, 16.03).63Race, n (%) .85African19 (30.2%)9 (29.0%)10 (31.3%) Mixed parentage44 (69.8%)22 (71.0%)22 (68.8%) Gender, n (%) .42Female55 (87.3%)26 (83.9%)29 (90.6%) Male8 (12.7%)5 (16.1%)3 (9.4%) >1 psychiatric diagnosis, n (%)35 (55.6%)18 (58.1%)17 (53.1%).69Mean no. of sessions, M (95% CI)9.46 (8.71, 10.00)9.55 (8.42, 10.00)9.38 (8.30, 10.00).73Treatment completers, n (%)52 (82.5%)25 (80.6%)27 (84.4%).95Expectancy score, M (95% CI)26.56 (25.30, 27.81)26.23 (24.46, 28.01)26.80 (25.00, 28.74).80 Table 1.Demographic variables of the intent-to-treat sample (N = 63).VariableTotal (N = 63)PE-A (n = 31)SC (n = 32)pAge, M (95% CI)15.35 (14.98, 15.72)15.19 (14.66, 15.72)15.50 (14.97, 16.03).63Race, n (%) .85African19 (30.2%)9 (29.0%)10 (31.3%) Mixed parentage44 (69.8%)22 (71.0%)22 (68.8%) Gender, n (%) .42Female55 (87.3%)26 (83.9%)29 (90.6%) Male8 (12.7%)5 (16.1%)3 (9.4%) >1 psychiatric diagnosis, n (%)35 (55.6%)18 (58.1%)17 (53.1%).69Mean no. of sessions, M (95% CI)9.46 (8.71, 10.00)9.55 (8.42, 10.00)9.38 (8.30, 10.00).73Treatment completers, n (%)52 (82.5%)25 (80.6%)27 (84.4%).95Expectancy score, M (95% CI)26.56 (25.30, 27.81)26.23 (24.46, 28.01)26.80 (25.00, 28.74).80 Demographic variables of the intent-to-treat sample (N = 63). Forty of the original 63 adolescents were available at the 60-month FU assessment: 20 from the PE-A treatment group, and 20 from the SC treatment group. Seven participants from the PE-A treatment group and eight of the SC treatment group reported experiencing a subsequent trauma. Trauma exposure included witnessing violence perpetrated against someone else (witnessing physical assault or a murder), experience of sexual assault or physical assault, and the death of a parent. None of the participants in this sample accessed additional psychological or pharmacological treatment. Adverse events During the study, adolescents who discontinued treatment did so of their own accord. All who discontinued had a final interview and did not accept alternative treatment or report any suicidal ideation. No serious adverse events were reported during the original randomised trial. Adolescents were also informed that their counsellors and the principal investigator could be contacted about any crisis outside of weekly treatment sessions. Treatment-emergent adverse events were documented in two participants who reported suicidal ideation which, upon further evaluation, was secondary to interpersonal conflict with a teacher and a parent, respectively. Suicide risk was assessed to be low for both adolescents and conflicts with the family and teacher were addressed. Both participants completed treatment without further incident. At the 24-month FU, two participants were identified with family or social problems unrelated to the trauma that had been the focus of treatment but required intervention. Both were referred to the appropriate intervention and followed up. Primary outcome Tables 2 and 3 depict the data from all assessment points for the primary outcome. Table 2.Primary outcomes at baseline and after treatment (means and standard deviations).Assessment timepointTotal (N = 63)PE-A (n = 31) SC (n = 32)Clinician-rated PTSD (CPSS-I) Baseline35.32 (7.85)34.48 (7.87)36.13 (7.86)Mid-treatment24.13 (12.67)22.08 (12.92)26.04 (12.36)Post-treatment12.37 (12.07)5.88 (7.37)18.28 (12.56) 3-month follow-up9.51 (11.13)4.96 (6.55)13.57 (12.81) 6-month follow-up9.29 (11.65)4.15 (5.90)14.64 (13.72) 12-month follow-up9.40 (10.60)5.04 (6.26)13.59 (12.24) 24-month follow-up10.02 (12.88)5.70 (7.83)14.17 (15.38) 60-month follow-up9.90 (11.07)8.60 (8.39)11.20 (13.32) Table 3.Effect sizes, confidence intervals and p-values for within and between group comparisons of CPSS-I Scores.Assessment timepointComparison typeGroup(s)Effect size (Hedges’ g)95% CIp-valueMIDWithin-groupPE-A1.17[0.61, 1.73]<.01 Within-groupSC0.94[0.40, 1.47]<.01 Between-groupPE-A vs SC0.39[−0.15, 0.93]=.15POSTWithin-groupPE-A2.72[1.99, 3.45]<.01 Within-groupSC1.70[1.10, 2.30]<.01 Between-groupPE-A vs SC1.19[0.60, 1.78]<.013-month follow-upWithin-groupPE-A2.81[2.07, 3.55]<.01 Within-groupSC2.12[1.49, 2.75]<.01 Between-groupPE-A vs SC0.84[0.28, 1.40]<.016-month follow-upWithin-groupPE-A2.88[2.14, 3.62]<.01 Within-groupSC2.08[1.43, 2.72]<.01 Between-groupPE-A vs SC0.98[0.40, 1.56]<.0112-month follow-upWithin-groupPE-A2.79[2.06, 3.52]<.01 Within-groupSC2.07[1.44, 2.70]<.01 Between-groupPE-A vs SC0.89[0.32, 1.45]<.0124-month follow-upWithin-groupPE-A2.79[2.04, 3.54]<.01 Within-groupSC2.06[1.41, 2.71]<.01 Between-groupPE-A vs SC0.89[0.30, 1.48]<.0160-month follow-upWithin-groupPE-A2.49[1.75, 3.23]<.01 Within-groupSC2.32[1.61, 3.04]<.01 Between-groupPE-A vs SC0.33[−0.28, 0.95]=.28 Table 2.Primary outcomes at baseline and after treatment (means and standard deviations).Assessment timepointTotal (N = 63)PE-A (n = 31) SC (n = 32)Clinician-rated PTSD (CPSS-I) Baseline35.32 (7.85)34.48 (7.87)36.13 (7.86)Mid-treatment24.13 (12.67)22.08 (12.92)26.04 (12.36)Post-treatment12.37 (12.07)5.88 (7.37)18.28 (12.56) 3-month follow-up9.51 (11.13)4.96 (6.55)13.57 (12.81) 6-month follow-up9.29 (11.65)4.15 (5.90)14.64 (13.72) 12-month follow-up9.40 (10.60)5.04 (6.26)13.59 (12.24) 24-month follow-up10.02 (12.88)5.70 (7.83)14.17 (15.38) 60-month follow-up9.90 (11.07)8.60 (8.39)11.20 (13.32) Primary outcomes at baseline and after treatment (means and standard deviations). Table 3.Effect sizes, confidence intervals and p-values for within and between group comparisons of CPSS-I Scores.Assessment timepointComparison typeGroup(s)Effect size (Hedges’ g)95% CIp-valueMIDWithin-groupPE-A1.17[0.61, 1.73]<.01 Within-groupSC0.94[0.40, 1.47]<.01 Between-groupPE-A vs SC0.39[−0.15, 0.93]=.15POSTWithin-groupPE-A2.72[1.99, 3.45]<.01 Within-groupSC1.70[1.10, 2.30]<.01 Between-groupPE-A vs SC1.19[0.60, 1.78]<.013-month follow-upWithin-groupPE-A2.81[2.07, 3.55]<.01 Within-groupSC2.12[1.49, 2.75]<.01 Between-groupPE-A vs SC0.84[0.28, 1.40]<.016-month follow-upWithin-groupPE-A2.88[2.14, 3.62]<.01 Within-groupSC2.08[1.43, 2.72]<.01 Between-groupPE-A vs SC0.98[0.40, 1.56]<.0112-month follow-upWithin-groupPE-A2.79[2.06, 3.52]<.01 Within-groupSC2.07[1.44, 2.70]<.01 Between-groupPE-A vs SC0.89[0.32, 1.45]<.0124-month follow-upWithin-groupPE-A2.79[2.04, 3.54]<.01 Within-groupSC2.06[1.41, 2.71]<.01 Between-groupPE-A vs SC0.89[0.30, 1.48]<.0160-month follow-upWithin-groupPE-A2.49[1.75, 3.23]<.01 Within-groupSC2.32[1.61, 3.04]<.01 Between-groupPE-A vs SC0.33[−0.28, 0.95]=.28 Effect sizes, confidence intervals and p-values for within and between group comparisons of CPSS-I Scores. The IE assessed PTSD severity scores (CPSS-I) over the 60-month post-treatment FU period, showed that both PE-A and SC groups experienced significant improvement in PTSD severity from baseline to post-treatment assessment and maintained the gains at 3-, 6-, 12-, 24- and 60 months post treatment. As reported previously, the PE-A treatment group experienced a significantly greater improvement in PTSD symptom severity than the SC treatment group at all assessment points up to the 60-month FU [F(7, 343) = 2.86, p < .01]. The between-group effect sizes (Hedges’ g) at each assessment point were large and consistently favoured the PE-A group through the 24-month follow-up: post-treatment (g = 1.19), 3-month follow-up (g = 0.84), 6-month follow-up (g = 0.98), 12-month follow-up (g = 0.89), and 24-month follow-up (g = 0.89). By the 60-month follow-up, however, the advantage of PE-A over SC in reducing PTSD symptom severity was no longer statistically significant (p = .28; g = 0.33). In considering the impact of further trauma exposure on participants in both groups (PE-A n = 7 of 20 participants, SC n = 8 of 20 participants), we found that further trauma exposure between the 24-month FU and the 60-month FU did not lead to significant symptom change [PE-A: (p = .19; g = 0.32) and (SC: p = .3; g = 0.25)] (Figure 2). Figure 2.CPSS-I comparative mean scores between PE-A and SC treatment groups.A line graph comparing CPSS dash I total scores over time for PE dash A and SC groups, both dropping from pre to post then remaining low.The figure shows a line graph of treatment outcome scores over time for 2 groups labeled PE dash A and SC. The horizontal axis is labeled Time and lists 8 points from left to right: Pre, Mid, Post, 3 Month, 6 Month, 12 Month, 24 Month, and 60 Month. The vertical axis is labeled CPSS dash I Total and ranges from negative 5 to 45 in increments of 5. Each group is plotted with a separate line and has small vertical error bars at every time point. The PE dash A group line starts near the top of the scale at Pre, drops steeply by Mid, and falls further by Post to a value just above 5. From Post through 24 Month, the PE dash A scores stay near the bottom of the scale with a slight dip and then a mild rise by 60 Month, where the value is still low relative to baseline. The SC group line also begins high at Pre, sits slightly above the PE dash A value, and declines by Mid and Post, but remains higher than the PE dash A values at each matching time. From Post to 24 Month, SC scores show a shallow increase and then a small decrease by 60 Month, staying well below baseline and consistently above the PE dash A trajectory. All data are approximate. CPSS-I comparative mean scores between PE-A and SC treatment groups. Secondary outcomes When comparing the presence of a PTSD diagnosis over the 60-month FU period, we found no significant group difference [Wald(6) = 5.58, p = .47]. The PE-A treatment group experienced a significantly greater reduction in the presence of a PTSD diagnosis than the SC treatment group at the post-treatment assessment (p = .02) and at 12-month FU (p < .01). The PE-A and SC groups experienced a significant reduction in the presence of a PTSD diagnosis at post-treatment assessment and maintained these changes in diagnoses at the 3-, 6-, 12-, 24- and 60 months FU assessments (p < .01 at all assessment points) (Table 4). Table 4.Remission from PTSD on the MINI-KID in treatment completers by treatment group at each assessment timepoint.Assessment timepointPE-A n/N (%)SC n/N (%)OR (PE-A vs. SC)95% CIpBaseline0/31 (0.0)0/32 (0.0)———Post-treatment20/25 (80.0)13/27 (48.1)4.18[1.09, 18.64].0233 months18/24 (75.0)13/26 (50.0)2.93[0.78, 12.14].0866 months23/26 (88.5)16/25 (64.0)4.19[0.87, 27.83].05212 months25/26 (96.2)14/27 (51.9)21.94[2.77, 1024.78]<.00124 months19/23 (82.6)14/24 (58.3)3.30[0.75, 17.55].11160 months14/20 (70.0)12/18 (66.7)1.16[0.24, 5.69]1.000Notes: Remission = a negative MINI-KID PTSD module at that assessment. Odds ratios (OR) are conditional maximum likelihood estimates with exact confidence intervals; values greater than 1 indicate higher odds of remission in PE-A group. All participants met diagnostic criteria at baseline by design, so baseline remission was not tested. p values are two-tailed Fisher's exact tests across the six follow-up assessments. Table 4.Remission from PTSD on the MINI-KID in treatment completers by treatment group at each assessment timepoint.Assessment timepointPE-A n/N (%)SC n/N (%)OR (PE-A vs. SC)95% CIpBaseline0/31 (0.0)0/32 (0.0)———Post-treatment20/25 (80.0)13/27 (48.1)4.18[1.09, 18.64].0233 months18/24 (75.0)13/26 (50.0)2.93[0.78, 12.14].0866 months23/26 (88.5)16/25 (64.0)4.19[0.87, 27.83].05212 months25/26 (96.2)14/27 (51.9)21.94[2.77, 1024.78]<.00124 months19/23 (82.6)14/24 (58.3)3.30[0.75, 17.55].11160 months14/20 (70.0)12/18 (66.7)1.16[0.24, 5.69]1.000Notes: Remission = a negative MINI-KID PTSD module at that assessment. Odds ratios (OR) are conditional maximum likelihood estimates with exact confidence intervals; values greater than 1 indicate higher odds of remission in PE-A group. All participants met diagnostic criteria at baseline by design, so baseline remission was not tested. p values are two-tailed Fisher's exact tests across the six follow-up assessments. Remission from PTSD on the MINI-KID in treatment completers by treatment group at each assessment timepoint. Notes: Remission = a negative MINI-KID PTSD module at that assessment. Odds ratios (OR) are conditional maximum likelihood estimates with exact confidence intervals; values greater than 1 indicate higher odds of remission in PE-A group. All participants met diagnostic criteria at baseline by design, so baseline remission was not tested. p values are two-tailed Fisher's exact tests across the six follow-up assessments. When comparing BDI scores in the PE-A and SC groups over the 60-month post-treatment FU period, we found no significant group difference [F (7, 338) = 1.92, p < .07]. Both PE-A and SC groups experienced significant improvement in depression severity from baseline to post-treatment-assessment and maintained gains at the 3-, 6-, 12-, 24- and 60 months FU assessments. The PE-A treatment group maintained significantly greater improvement in depression symptom severity scores than the SC treatment group at the post-treatment assessment (p < .01; g = 0.88), at 3-month FU (p = .02; g = 0.61), at 6-month FU (p = .02; g = 0.50) and at 24-month FU (p = 0.03; g = 0.26), but not at 12-month (p = .17; g = 0.38) or at 60-month (p = .6; g = 0.16) FU. There was no significant group difference in improvement in functioning (CGAS scores) [F(7, 346) = 0.95, p = .47]. Both PE-A and SC groups experienced significant improvement in general functioning from baseline to post-treatment-assessment and maintained these gains at 60 months. Data for secondary measures are available in the Supplementary file. There was no significant difference in perceived treatment helpfulness (STO) between the treatment groups [F(6, 285) = 1.30, p = .26]. Adolescents in both groups continued to indicate that they found the treatment to be very helpful (mean = 29.91 out of 32) in terms of treating their PTSD, reducing other personal problems, improving confidence to recommend the treatment to others, and how logical the treatment seemed to them throughout the follow-up period. There was no significant difference in participants’ rating of the therapeutic alliance (WIA-CA) between the two groups [F(5, 232) = 1.06, p = .38]. Ratings of the therapeutic alliance improved significantly over time for both treatment groups [F(5, 232) = 2.91, p = .01]. There was no significant difference in general anxiety scores (MASC) between the groups [F(97, 333) = 0.37, p = .92]. Both treatment groups experienced a significant improvement from baseline and maintained these gains from post-treatment through the 60-month FU assessment. The total score improved from 65.23 and 70.94 at baseline in the PE-A arm and the SC arm, respectively, to 38.06 and 34.85 at the 60-month FU assessment. For comparison, the mean score for adolescent girls in one normative Swedish sample was 56 (Ivarsson, 2006). There was no significant difference in anger scores (STAXI-2) between the groups [F(7, 340) = 1.53; p = .15], with both groups experiencing significant improvement from baseline and maintaining these gains at 60 months. There was no significant difference in self-esteem scores (RSES) between the groups [F(7, 343) = 0.41; p = .90], or in resilience scores (CD-RISC) [F(7, 377) = 0.92; p = .49]. Both treatment groups experienced significant improvement from baseline and maintained these gains from the post-treatment to 60 months assessments. Furthermore, there was no significant between-group difference in emotion regulation scores (NMR) [F(7, 337) = 0.34; p = .94], with both groups experiencing significant improvement from baseline and maintaining these gains until the 60-month FU. Finally, we documented no significant between-group difference in social support scores (MSPSS) [F(7, 337) = 0.70; p = .67]. Discussion Our findings of sustained treatment gains up to 60 months post-intervention, together with previously reported 24 months follow-up data in an adolescent sample diagnosed with PTSD (Rossouw et al., 2018, 2022), make an incremental contribution to the relatively limited literature base on the long-term impact of task-shifted, evidence-based treatments in LMICS. Inclusion of previously reported follow-up data in this paper serves to illustrate the symptom trajectories over time of PTSD and co-occurring phenomenology. Treatment gains after receiving either PE-A or SC were durable in both treatment arms at 5 years post-treatment. Notably, adolescents in the PE-A treatment group experienced significantly greater reduction in PTSD symptoms than those in the SC treatment group at all FU assessment points, except at the 60-month timepoint where there was no significant between-group difference. It is possible that the loss of some participants (a further 7 participants in total; 3 in the PE-A group and 4 in the SC group) between the 24-month FU and the 60-month FU, due to the adolescents graduating and moving away, may have reduced the statistical power for analyses, thereby affecting the robustness of the findings. Importantly, subsequent trauma exposure between the 24- and 60-month follow-up assessments did not result in a significant change in PTSD symptom severity in either treatment arm. The finding that PTSD symptom severity remained stable in both treatment groups between the 24-month and 60-month follow-up assessments suggests that the therapeutic gains achieved were sustained over time and may indicate that treatment for PTSD confers resilience against the potential impact of subsequent trauma exposure. On all other secondary measures, the groups did not differ significantly. The issue of palatability and tolerability of exposure-based therapies and the literature on dropout rates from TF-CBT interventions has been inconsistent. However, a meta-analysis about the palatability of these interventions in youth has shown that dropouts from RCTs are not more likely for trauma-focused treatments than for non-trauma-focused arms or control conditions (Simmons et al., 2021). This was consistent with the results of our original RCT (Rossouw et al., 2018), where the number of participants who did not complete the full course of treatment and were considered dropouts did not differ between the two treatment arms. This study demonstrates that in a LMIC setting, treatment can be task-shifted to nurses without prior psychotherapy experience. When asked to rate the helpfulness of the treatments and the therapeutic alliance with their counsellor, adolescents in both treatment arms rated both treatments to be very helpful to them and the therapeutic alliance with their counsellor to be excellent. The self-reported measures aligned with qualitative data obtained from participants who highlighted the ease of access to treatment in a school-based intervention, the provision of treatment by providers from outside their community that addressed concerns around community-based stigma related to accessing mental health treatment, and concerns about confidentiality and the warmth and caring displayed by treatment providers and recruiters (Van der Water et al., 2017; Van der Water, Rossouw, van der Watt et al., 2018; Van der Water, Rossouw, Yadin et al., 2018). The improvement in global functioning observed in both treatment arms at the 24-month and 60-month post-treatment assessments represents a noteworthy positive outcome, particularly given that many adolescents experienced additional traumatic events during the interval between these follow-up assessments. Despite that, adolescents in both treatment arms reported experiencing a significant improvement in functioning at the 60-month FU assessment, extending previously reported results (Rossouw et al., 2022). The long-term assessment results provided support for the hypothesis of the study that both treatment groups would maintain treatment gains over time. The hypothesis that PE-A would retain its superiority over SC in reducing both interviewer-assessed PTSD severity was supported up to the 24-month follow-up point. At the 60-month follow-up assessment, with a smaller reachable sample size (n = 40) of the original study participants (n = 63), the superiority of the PE-A treatment group over the SC group in PTSD symptom severity was no longer statistically significant. Limitations The original RCT was conducted with participants in school, which contributed to the compliance with the assessments. By the 60-month follow-up assessment, many participants had transitioned into new phases of their lives and had become geographically dispersed, making them more difficult to locate and reassess. This contributed to attrition from the original cohort, resulting in a reduced sample size that may have limited the statistical power of the analyses, influenced the observed findings, and reduced the generalisability of the results. A further limitation relates to the setting in which the interventions were delivered. As treatment was provided within schools, the findings may not readily generalise to other community-based service settings, such as primary healthcare clinics or specialist mental health services, where implementation contexts, resources, and patterns of engagement may differ. Future studies These data demonstrating that established, evidence-based treatments can be effectively task-shifted and implemented with adolescents from LMICs with long-term enduring positive results are encouraging. Future studies should identify ways to scale this up to larger and more diverse adolescent populations. The inevitable reality is that young people in South Africa are likely to disproportionally encounter further trauma exposure in their lifetime. However, the preliminary information from this study that subsequent experience of traumatic events did not seem to impact the maintenance of gains made in the original treatment phase may be an important finding. Further research warrants asking the question whether treatment for PTSD in youth, and in particular trauma-focused treatment such as PE-A, has a protective effect when a person encounters a subsequent trauma. Conclusion The current data, together with the results of the original RCT, showed that the positive therapeutic effect on PTSD symptoms of a relatively short treatment protocol (mean number of 9 sessions of either PE-A or SC), task-shifted to previously psychotherapy-naïve nurses in a LMIC setting with limited mental health resources, was durable upto 5 years post-treatment. This documented long-term maintenance of a significant reduction in PTSD symptom severity, with sustained improved levels of functioning, provides a strong case for the upscaling of both treatments, and of PE-A in particular, within a community-based context. Supplementary Material Supplementary Material Final.docx