Participants were male soldiers with a deployment-related PTSD. Female participants were not included by methodological design consideration, as the study involved the assessment of biological stress markers (e.g., cortisol), and menstrual cycle-related hormonal variability would have introduced additional potential bias. Soldiers were included into the study if they met the following inclusion criteria: (1) age > 18 years, (2) diagnosed PTSD, and (3) fluency in written and spoken German language, as well as unimpaired vision and hearing. Exclusion criteria were: (1) comorbid psychosis, neurological illnesses, and severe medical disorders, (2) the use of highly potent neuroleptics, and (3) acute suicidality. Comorbid depression, anxiety disorder, and adjustment disorder were no exclusion criteria. The soldiers were recruited at the military hospitals in Berlin and Ulm. All study participants received inpatient treatment with trauma-focused therapy in military hospitals. The study was approved by the ethics committee of Witten/Herdecke University (Application No. 196/2016) and is in compliance with the Declaration of Helsinki. A total of N = 53 participants were included in the study. Eligible soldiers were identified and invited to participate in the context of routine clinical care. Only soldiers who provided written informed consent for study participation were included systematic documentation. The total number of potentially eligible soldiers who were approached but declined participation was not recorded; therefore, a formal participation rate cannot be reported. In accordance with German data protection regulations, documentation of potential participants was only permitted after written informed consent had been obtained.

In this study with group and pre- and post-intervention comparisons, soldiers with PTSD were quasi-randomly assigned to one of two groups, which does not fully control for expectancy effects or pre-existing group differences. The IG received an app-based neuropsychological training developed by the company HeadApp (Gommern, Germany; https://www.headapp.com/de/startseite-2-2/). The c-bnt program comprised modules targeting attentional control, working memory updating, planning, and cognitive flexibility. These processes were hypothesized to indirectly support long-term verbal and visuospatial memory performance by facilitating efficient encoding, strategic organization, and executive monitoring during learning and retrieval. However, the intervention was not specifically designed to train long-term verbal or visuospatial memory functions, and transfer to standardized long-term memory measures was therefore considered exploratory. The tauG was a treatment as usual group treated with trauma therapy only. All participants in both groups received inpatient trauma-focused treatment as usual (TAU) at German military hospitals. TAU was delivered by licensed trauma therapists and comprised different evidence-based trauma-focused approaches, including Eye Movement Desensitization and Reprocessing (EMDR), Imagery Rescripting and Reprocessing Therapy (IRRT), and Cognitive Processing Therapy (CPT). Trauma-focused psychotherapy was typically delivered at a frequency of approximately two sessions per week, consistent with standard inpatient military care. In line with standard inpatient military care, TAU followed a multimodal treatment approach and additionally included complementary therapeutic elements (e.g., psychoeducation, physical activity, and supportive interventions), which were equally available to both groups. Assignment of patients to therapists occurred randomly as part of routine clinical care. Treatment was not manualized within the framework of the present study, and specific therapeutic modalities were not documented at the individual level. All participants underwent a comprehensive clinical and neuropsychological examination at baseline (t₁) and after three weeks of therapy (t₂). The study design is illustrated in Figure 1. Of the N = 53 participants assessed at baseline (t₁), n = 34 completed the post-intervention assessment at t₂, resulting in an attrition rate of 35.85%.

Independent samples t-tests at t₁ revealed no significant baseline differences between the intervention and treatment as usual groups on any PCL subscale (see Table 2a): Intrusion (t(42) = 0.63, p = .529, d = 0.19), Avoidance (t(42) = –0.08, p = .936, d = 0.02), Negative Cognitions (t(42) = –0.84, p = .408, d = 0.25), Arousal (t(42) = –0.08, p = .935, d = 0.03), and overall PTSD symptom severity (t(42) = –0.22, p = .831, d = 0.07). There was also no significant baseline difference in depressive symptom severity (BDI-II) between the IG and tauG (t(41) = –1.07, p = .291, d = 0.327).

Independent-samples t-tests were conducted to examine whether post-treatment (t₂) symptom levels were lower in the IG compared to the tauG (see Table 2b). Significant group differences emerged for Avoidance (t(32) = –1.84, p = .038, d = 0.63), Negative Cognitions (t(32) = –1.76, p = .044, d = 0.60), and depressive symptoms (t(32) = –2.03, p = .026, d = 0.69), all indicating moderate effect sizes. No significant differences were found for Intrusion (p = .283, d = 0.32), Arousal (p = .094, d = 0.47), or overall PTSD severity (p = .074, d = 0.49), though small to moderate effects were observed.

All statistical assumptions (normality, homogeneity of variance, and equality of covariance matrices) were assessed and considered in the interpretation of results. Mixed-design ANOVAs revealed no significant time × group interactions for any of the PTSD symptom domains: Intrusion, F(1, 30) = 0.70, p = .411, η² = .023; Avoidance, F(1, 30) = 0.42, p = .520, η² = .014; Negative Cognitions, F(1, 30) = 0.16, p = .688, η² = .005; and Arousal, F(1, 30) = 0.28, p = .603, η² = .009. Due to violations of variance homogeneity, the interaction effects for Arousal and overall PTSD symptom severity could not be interpreted.

Regarding the main effects of time, no significant changes in symptom severity were found across time points for Intrusion (p = .828), Negative Cognitions (p = 1.000), Arousal (p = 1.000), or total PTSD symptom severity (p = .753). A marginal trend was observed for Avoidance, F(1, 30) = 3.28, p = .080, η² = .098.

Similarly, no statistically significant main effects of group were observed for any domain: Intrusion, F(1, 30) = 0.09, p = .767, η² = .003; Avoidance, F(1, 30) = 3.66, p = .065, η² = .109; Negative Cognitions, F(1, 30) = 3.89, p = .058, η² = .115; Arousal, F(1, 30) = 1.49, p = .232, η² = .047; and total PTSD symptom severity, F(1, 30) = 2.29, p = .141, η² = .071. Table 3 summarizes all statistical parameters of the mixed-ANOVAs for PTSD symptom clusters, including interaction, time, and group effects, along with associated effect sizes. Note that interaction effects for Arousal and total PTSD symptoms were not interpreted due to significant variance inequality.

For the BDI II measures of depression, all statistical assumptions were met. A mixed-design ANOVA revealed no significant time × group interaction, F(1, 29) = 0.06, p = .811, η² = .002, indicating that changes in depressive symptoms over time did not differ between the intervention and control groups. The main effect of time was not significant, F(1, 29) = 0.17, p = .688, η² = .006, suggesting no overall change in depressive symptoms from pre- to post-assessment. The main effect of group was also not statistically significant, F(1, 29) = 3.44, p = .074, η² = .106, indicating no substantial overall difference in depressive symptom levels between groups.

Group comparisons concerning memory performance between the IG and the tauG at t₁ independent samples t-tests revealed no significant baseline differences on any VLMT trial: A1–A5 (t(49) = 0.17, p = .869, d = 0.05), A6 (t(49) = –0.11, p = .915, d = –0.03), and A7 (t(49) = –0.82, p = .418, d = –0.23).

Similarly, no significant group differences at t₁ were found on any VWLT subscale (see Table 4a): A1–A5, t(42) = 1.32, p = .193, d = 0.39; B, t(42) = 1.21, p = .233, d = 0.36; A6, t(42) = 1.97, p = .055, d = 0.60; A7, t(42) = 1.19, p = .240, d = 0.36; forgetting rate, t(42) = 0.70, p = .486, d = 0.21. Effect sizes were small to moderate, with the largest (yet nonsignificant) effect for short-delay recall (A6), indicating a trend toward better performance in the IG.

To assess potential group differences in verbal learning and memory performance at post-assessment (t₂), independent samples t-tests were conducted for the VLMT subscales A1–A5, A6, and A7. No significant differences were found between the intervention and treatment as usual groups on any of the subscales for learning performance (A1–A5), t(37) = 0.18, p = .429, d = 0.06, short-delay- recall (A6), t(36.94) = −0.13, p = .450, d = −0.04, and long-delay recall (A7), t(32.72) = −0.63, p = .265, d = −0.20. Effect sizes were negligible to small across all subscales.

No significant group differences were found for learning performance (t(34) = 0.36, p = .357, d = 0.12), interference performance (t(34) = 0.16, p = .435, d = 0.06), short-delay recall (t(34) = 0.65, p = .261, d = 0.22), or long-delay recall (t(34) = 0.01, p = .495, d = 0.01). All effects were small and not statistically significant. However, for forgetting rate, a trend-level group difference emerged in favor of the tauG (t(34) = −1.56, p = .064, d = −0.52), indicating a lower forgetting rate in the tauG at t₂. The IG, in contrast, showed a poorer outcome on this measure. The effect size was in the medium range, and the result approached significance in the expected direction. An overview of group differences at t₂ is provided in Table 4b.

Before conducting the mixed-design ANOVAs, assumptions were tested. Levene’s and Box’s M tests showed no violations of homogeneity of variances or equality of covariance matrices. Shapiro-Wilk tests confirmed normality for learning trials (A1–A5) at both time points. In contrast, deviations from normality were found in the recall conditions. A6 (direct recall) deviated significantly in the IG at t₁ (p = .014) and in both groups at t₂ (p <.05). For A7 (delayed recall), the tauG deviated at t₁ (p = .005), and both groups at t₂ (p <.05). These results indicate moderate normality violations for direct and delayed recall, warranting cautious interpretation.

For VLMT A1–A5, there was no significant time × group interaction, F(1, 37) = 1.16, p = .288, η² = .030. No significant main effects of time, F(1, 37) = 1.06, p = .309, η² = .028, or group, F(1, 37) = 0.04, p = .853, η² = .001, were observed.

For VLMT A6, the analysis also revealed no significant time × group interaction, F(1, 37) = 1.50, p = .228, η² = .039, and no significant main effects of time, F(1, 37) = 0.07, p = .787, η² = .002, or group, F(1, 37) = 0.29, p = .593, η² = .008.

In contrast, the analysis for VLMT A7 revealed a statistically significant time × group interaction, F(1, 37) = 8.51, p = .006, η² = .187, reflecting a large effect size. No significant main effect of time, F(1, 37) = 1.27, p = .267, η² = .033, or group, F(1, 37) = 0.09, p = .762, η² = .003, was found. Descriptive statistics for all VLMT scales are presented in Table 7a.

The Shapiro–Wilk test revealed significant deviations from normality for Interference Performance (B) in both groups at both time points, for Short-Delay Recall (A6) in the IG at t₂, and for the Forgetting Rate in the tauG at t₂ (p < .05). All other variables met the assumption of normality. Levene’s test confirmed homogeneity of variances, except for Long-Delay Recall (A7) at t₁ (p = .023). Sphericity was confirmed for all within-subject comparisons (Mauchly’s W = 1.000).

For learning performance (A1–A5), neither main nor interaction effects were significant (F(1, 34) = 1.50, p = .230, η² = .042; interaction: F(1, 34) = 2.01, p = .165, η² = .056), indicating stable performance with small to medium effects.

Interference performance (B) also remained unchanged over time (F(1, 34) = 1.01, p = .323, η² = .029), and the interaction effect was nonsignificant (F(1, 34) = 1.48, p = .232, η² = .042), reflecting small effects.

For short-delay recall (A6), the main effect was nonsignificant (F(1, 34) = 0.01, p = .945, η² = .000), but a marginally significant interaction emerged (F(1, 34) = 3.51, p = .070, η² = .094), suggesting a potential medium effect and differential group development.

Long-delay recall (A7) showed a significant interaction (F(1, 34) = 5.46, p = .025, η² = .138), indicating a medium-to-large group effect with greater gains in the tauG. The main time effect was nonsignificant (F(1, 34) = 1.98, p = .168, η² = .055).

Finally, the forgetting rate (A5–A7) showed a significant main effect of time (F(1, 34) = 8.38, p = .007, η² = .198), which constitutes a large effect, indicating that forgetting significantly increased across sessions. The group × time interaction showed a statistical trend (F(1, 34) = 3.37, p = .075, η² = .090), suggesting a medium effect and possible group-specific differences in memory consolidation trajectories. An overview of all interaction, time, and group effects from the repeated-measures ANOVAs for the VLMT and VWLT subscales is presented in Table 5.

The following analyses are exploratory and are reported to provide a comprehensive overview of the data; given the limited sample size and statistical power, null findings should be interpreted with caution.

To examine whether depressive symptoms mediate the link between posttraumatic stress and verbal memory, three mediation analyses were run using PROCESS (Model 4, 5,000 bootstraps, 95% CI). PTSD symptoms at baseline (PCL t₁) served as the predictor (X), depressive symptoms at post-assessment (BDI-II t₂) as mediator (M), and verbal memory outcomes at post-assessment as dependent variables (Y): verbal learning (VLMT A1–A5), short-delay recall (VLMT A6), and long-delay recall (VLMT A7).

For verbal learning (A1–A5 t₂), the total effect of PTSD symptoms was nonsignificant, b = –0.111, p = .242. PTSD symptoms predicted depressive symptoms, b = 0.603, p <.001, but depressive symptoms did not predict verbal learning, b = 0.004, p = .985. The indirect effect was nonsignificant, b = 0.002, 95% CI [–0.220, 0.211] (see Figure 3a).

For direct recall (A6 t₂), the total effect of PTSD symptoms was nonsignificant, b = –0.014, p = .652. PTSD symptoms predicted depressive symptoms (b = 0.603, p <.001), but depressive symptoms did not predict recall (b = 0.029, p = .638). Neither the direct effect (b = –0.031, p = .457) nor the indirect effect (b = 0.018, 95% CI [–0.050, 0.089]) was significant (see Figure 3b).

For delayed recall (A7 t₂), the total effect of PTSD symptoms was nonsignificant, b = – 0.018, p = .529 (see Figure 3c). PTSD symptoms predicted depressive symptoms (b = 0.603, p <.001), but depression did not predict delayed recall (b = 0.017, p = .782). The indirect effect was nonsignificant, b = 0.010, 95% CI [–0.053, 0.082].

Across all models, PTSD symptoms predicted higher depressive symptoms, but depression did not significantly predict verbal memory outcomes. Thus, no evidence emerged for depression mediating the relationship between PTSD symptoms and verbal memory performance. Effect estimates are summarized in Table 6.

To test whether depressive symptoms mediate the link between PTSD severity and visuospatial memory, mediation analyses (PROCESS Model 4, 5,000 bootstrap samples) were conducted. PCL (t₁) served as the predictor, BDI-II (t₂) as the mediator, and VWLT subscales at t₂ as outcomes: learning (A1–A5), interference (B), short-delay recall (A6), long-delay recall (A7), and forgetting rate (A5–A7). Across all models, PCL consistently predicted BDI-II (B = 0.56, p <.001), but neither the direct nor indirect effects on VWLT outcomes were significant, as all 95% confidence intervals included zero. Thus, no evidence emerged for a mediating role of depression in visuospatial memory performance.

Effect estimates are summarized in Table 7, and the model structure is visualized in Figures 4a-e, illustrating the consistent model and nonsignificant mediation effects.

Our results are in accordance with previous studies showing that cognitive training can improve PTSD-related symptoms. Meta-analyses have reported small to moderate symptom reductions after working memory training and interventions targeting attention bias modification in PTSD populations (38, 39). These approaches also led to clinical improvement without robust enhancement of cognitive functions.

However, our results are only partially consistent with previous studies that identified robust verbal and visuospatial memory impairments in PTSD populations (34, 35). For example, Johnsen and Asbjørnsen (29) reported large deficits in verbal memory compared to healthy controls in civilian PTSD samples, and Sep, Geuze (35) documented broad impairments in episodic memory across both clinical (mostly civilian) and preclinical PTSD samples. Our sample, comprising active-duty military personnel undergoing inpatient treatment, did not exhibit clear memory deficits at t₁ baseline measures. VLMT scores were comparable to respective age norms. Visuospatial memory performance (VWLT) was only slightly below normative age average. Thus, our data do not support a pronounced memory dysfunction in this clinical inpatient sample.

However, direct comparability between civilian and military PTSD samples is limited. Civilian PTSD samples typically show a substantially higher proportion of complex PTSD (cPTSD), which is characterized by a higher prevalence of sequential or prolonged traumatic exposure, more pronounced negative cognitions, dissociative symptoms, and marked impairments in emotion regulation. These features are known to be associated with broader and more persistent cognitive dysfunction, including deficits in memory and executive functioning. Accordingly, differences in long-term memory performance between civilian and military PTSD samples may partly reflect differences in trauma structure and symptom profiles rather than true inconsistencies across studies. In addition, military populations differ from civilian samples with respect to premorbid cognitive functioning and selection processes. Military service involves cognitive, physical, and psychological selection mechanisms, and prospective studies have shown largely preserved neuropsychological functioning in soldiers following deployment, particularly in attention and executive domains (53). Such premorbid cognitive resources may buffer against the development of pronounced long-term memory deficits, even in the presence of trauma exposure, and may further contribute to differences in cognitive profiles between military and civilian PTSD populations. In addition, the nature of military trauma differs in important ways from most civilian traumata. Combat-related trauma is often characterized by repeated exposure and may involve moral or existential conflict; however, the assumption of a simple cumulative dose-response relationship remains controversial. Evidence suggests that symptom severity and cognitive outcomes are not solely determined by the number of deployments, but are strongly influenced by the subjective meaning, intensity, and moral context of traumatic experiences (54). Future studies should therefore investigate how individual predispositions, trauma type, and military-specific experiences interact to shape neurocognitive outcomes in PTSD.

Few studies have specifically investigated training effects on long-term memory in PTSD. Moradi, Moshirpanahi (33) applied Memory Specificity Training in combat veterans and showed improvements in autobiographical memory and reductions in PTSD symptoms. Bomyea, Caudle (32) tested a computerized working memory intervention in U.S. veterans and report symptom reductions. Although participants slightly improved within the training environment due to repeated practice and feedback, these improvements did not generalize to untrained standardized neuropsychological memory tests. Our data are in line with these previous reports in that we also found symptom reduction without robust improvement of verbal and visuospatial long-term memory.

The lack of clear memory enhancement in our sample may at least in part be due to relatively intact long-term memory functions at t₁ in both groups. Moreover, the relatively short training duration and the high therapeutic load during inpatient trauma treatment as usual may have diluted significant effects on untrained long-term memory tasks in the IG. It should be noted that the c-bnt intervention was not specifically designed to target long-term memory consolidation processes. Any potential effects on verbal or visuospatial long-term memory were considered indirect, for example via general task engagement or attentional demands, rather than through targeted modulation of emotion–cognition integration or memory consolidation processes. Given the intervention duration of three weeks in the absence of targeted consolidation training, measurable changes in standardized long-term memory performance were not hypothesized. In addition, baseline performance on the VLMT and VWLT was largely within the normative range, which may have limited observable improvement due to ceiling effects. The lack of observable transfer from the trained to untrained tasks in our study may suggest that the cognitive resources targeted in the training may not have sufficiently overlapped with those required in standard neuropsychological long-term tests such as the VLMT and VWLT. In addition, the relatively high training intensity of approximately one hour per day, delivered in parallel with intensive inpatient trauma-focused therapy, may have contributed to cognitive fatigue or overload. Trauma-focused treatment places substantial emotional and cognitive demands on patients, and additional daily cognitive training may have taxed available cognitive resources rather than facilitating consolidation. Such overload effects could have attenuated potential benefits of the training on long-term memory performance, particularly in domains requiring sustained encoding and consolidation processes. This highlights a central challenge in cognitive intervention research. For training-induced improvements to generalize, the trained tasks need to engage core processes shared with transfer tasks, a condition that may not have been sufficiently met in the present intervention. Future studies should therefore focus on the degree to which specific cognitive functions trained in each module may concretely lead to neurocognitive improvements in broader, functionally related domains.

Our finding that depressive symptoms could be reduced by c-bnt, but depression did not mediate memory performance, contrasts with prior evidence linking depression to cognitive dysfunction. Clinical studies and meta-analyses have demonstrated that individuals with depression, especially in clinical and inpatient samples, show consistent impairments in executive functioning (e.g., cognitive flexibility, inhibitory control), attention regulation, and episodic memory, particularly in verbal learning and delayed recall (19, 23, 24). Military-specific studies further indicate that depression is associated with cognitive inefficiencies. In a large veteran sample, depressed individuals exhibited significant impairments in declarative long-term memory and attention, suggesting that memory-related difficulties are also relevant in military populations (55). In addition, neuroscientific research has shown that depression can impair the encoding and retrieval of episodic memories, particularly for positive stimuli, through dysregulation of reward-related neural systems (56).

However, in our sample, the reduction in depressive symptoms after c-bnt did not translate into improved verbal or visuospatial long-term memory. This suggests that depressive symptoms alone do not explain the cognitive deficits commonly found in PTSD, which may instead be rooted in broader neurocognitive dysfunctions. Supporting this view, Scott, Lynch (5) demonstrated that memory and executive abilities independently predicted treatment outcomes in veterans with comorbid PTSD and substance use disorders, highlighting that cognitive functioning itself is a critical factor in determining therapeutic success.

Importantly, we observed symptom reductions in both PTSD and depression, yet memory performance remained mainly unchanged. This finding is consistent with previous research suggesting that cognitive impairments in PTSD, particularly in verbal and visuospatial long-term memory, may be relatively stable and require long and intensive interventions to produce substantial improvements (26, 35). For example, Schweizer, Samimi (38) demonstrated moderate improvements in cognitive control and affect regulation following a 6-week emotional working memory training in adolescents with PTSD. Likewise, Bomyea, Caudle (32) found significant symptom reductions after a 4-week computerized working memory training in U.S. veterans, but only minimal gains on untrained neuropsychological tasks. Polak, Witteveen (26) concluded that executive and memory functions in PTSD are often resistant to change, especially in short-term interventions. Given the data, the relatively brief duration of our c-bnt (three weeks) may have been too short to elicit observable transfer effects on standardized long-term memory measures. The apparent stability of cognitive performance in our sample likely reflects the chronic nature of PTSD-related memory deficits and illustrates the therapeutic challenge of achieving cognitive improvements in chronic cases.

Although our intervention was delivered in parallel with standard trauma-focused inpatient treatment, c-bnt itself did not include trauma-specific therapeutic components such as narrative exposure, imaginative reliving, or cognitive restructuring. Instead, it was a pragmatic and practical cognitive training, consisting of task-oriented training modules targeting attention, planning, and memory through repetitive, emotionally neutral computer-based exercises. Therefore, the observed reductions in avoidance and negative cognitions in the IG may not result from progress in trauma processing per se, but rather from indirect therapeutic mechanisms. Although the training did not explicitly target emotional processing, several task characteristics may nonetheless be relevant for emotion-cognition integration. The c-bnt modules required sustained attention, executive monitoring, and adaptive adjustment to increasing task demands. Such processes are closely linked to cognitive control mechanisms that support emotion regulation under conditions of stress or heightened arousal. From this perspective, c-bnt may have indirectly engaged emotion-cognition integration and coordination processes, providing a plausible, though indirect, mechanistic link to reductions in avoidance and negative cognitions. Empirical findings suggest that emotion-focused cognitive trainings, such as emotional working memory training (44), can enhance affective control and attentional regulation without incorporating trauma-focused psychotherapeutic interventions. For instance, Schweizer, Grahn (44) demonstrated that emotional working memory training improved PTSD symptoms by strengthening top-down control and affect regulation. These effects were attributed to increased cognitive activation and reinforcement of self-regulatory capacities.

Moreover, structured task engagement, such as implemented in our c-bnt modules, may foster a sense of mastery and self-efficacy, psychological resources known to buffer against depressive symptoms and improve psychological functioning (5). Such indirect mechanisms may have contributed to symptom improvements observed in our study in the absence of significant cognitive gains or mediation by depression. Furthermore, the regular structure and positive feedback within the training modules may have acted as behavioral activation elements, which are known to alleviate depressive symptoms. These active components of the applied c-bnt may have fostered self-efficacy in the soldiers. Empirical evidence highlights the importance of perceived self-efficacy in veterans. Blackburn and Owens (57) found that general self-efficacy was significantly associated with lower PTSD and depression severity among veterans. The authors argue that self-efficacy may successfully buffer the impact of combat exposure. In a sample of Iraq/Afghanistan veterans, higher self-efficacy was linked to a lower likelihood of seeking mental health treatment. This finding confirms the important role of self-efficacy for mental health of soldiers in that it demonstrates its link to self-regulatory coping and resilience (58). The structured, mastery-driven nature of c-bnt, characterized by repeated task success and positive feedback, may thus have enhanced self-efficacy and indirectly contributed to depression and PTSD symptom improvement in our study. Benefits in self-efficacy may also explain reduction of depressive symptoms in the absence of statistically significant gains in long-term memory performance and mediating effects of depression. At the same time, because both groups received identical inpatient trauma-focused treatment, additive effects of c-bnt cannot be fully disentangled from nonspecific factors. Increased engagement in structured daily activities, additional cognitive stimulation, and heightened task involvement, reflected in sustained attentional engagement, repeated goal-directed task execution, and continuous performance feedback, may have contributed to symptom reduction independently of specific training effects. Moreover, expectancy effects related to receiving an additional intervention may have influenced outcomes, particularly in the absence of blinding of the study. These nonspecific mechanisms should therefore be considered as potential contributors to the observed effects and warrant more stringent control in future studies. Depressive symptoms and PTSD-related negative cognitions represent closely related, but distinct constructs, which is reflected in their strong conceptual and empirical association across analyses. Future studies should therefore investigate differential effects of specific c-bnt modules and tasks on mediators like self-efficacy.

This study has several methodological and conceptual strengths. First, it represents one of the few controlled trials to examine the effects of c-bnt on both clinical symptoms and cognitive functioning in PTSD. While previous research has predominantly addressed single cognitive domains such as working memory or attention (38, 39), our study broadened the focus to include both verbal and visuospatial long-term memory functions. This approach aligns with the growing evidence that PTSD affects multiple memory systems (35) and responds variably to cognitive interventions. Second, the study applied a theory-driven mediational framework, testing whether depressive symptoms could explain potential links between PTSD severity and memory performance. Although the mediation models did not yield significant indirect effects, we found a positive association between PTSD symptom severity at baseline and depressive symptom severity at post-assessment. This relationship supports the assumption of interrelated symptom domains and justifies the application of a mediational framework, even in the absence of a statistically significant mediation effect. Third, the study used well-established, standardized neuropsychological assessments (VLMT, VWLT) to objectively measure cognitive functions. Moreover, both verbal and visuospatial domains were assessed using parallel learning and recall paradigms, allowing for a differentiated analysis of modality-specific effects. Fourth, the study targeted a military population, a group that is both clinically relevant and underrepresented in cognitive training research. PTSD in military contexts often presents chronic trajectories, high functional impairment, and complex comorbidities, making this group an important clinical target for non-trauma-focused interventions. Finally, the study applied a c-bnt that could be implemented flexibly in clinical or occupational settings. This increases the ecological validity of the approach and points to practical applications beyond specialized PTSD clinics. Although the intervention required daily training time in addition to trauma therapy, its standardized, self-guided, and non-trauma-focused format may still offer practical value as a low-threshold adjunct in settings where access to psychotherapy is limited or where patients are reluctant to engage in trauma-focused interventions. The standardized and fully digital format reduces therapist dependency and allows for individualized training schedules, which is particularly relevant for military personnel with irregular schedules or operational deployments. Moreover, by engaging participants in structured, self-directed cognitive exercises, the intervention may foster a sense of self-efficacy and active involvement in the therapeutic process, an aspect that is often limited in traditional trauma-focused treatments, which tend to be more therapist-centered.

The findings of this study suggest that c-bnt may serve as a pragmatic, low-threshold adjunct to trauma-focused therapy in military settings that may contribute to modest, domain-specific symptom improvement in PTSD and depression. Improvement of avoidance and negative cognitions is of particular clinical relevance in military populations, where emotional numbing, avoidance behavior, and functional impairment can compromise mission readiness, interpersonal functioning, and reintegration into social and occupational roles. Moreover, a reduction in avoidance may not only support daily coping but also therapeutic engagement, which is often hindered by stigma, distrust, or fear of reactivation in active-duty personnel.

For soldiers who refuse psychotherapy at all or do not want to engage in trauma-focused treatment approaches, c-bnt may offer a preparatory strategy of stabilization. By activating cognitive processes through emotionally neutral and mastery-oriented tasks, it may reduce arousal and foster psychological readiness for subsequent trauma processing. Given that access to psychotherapy is often limited due to deployment schedules, geographic dispersion, or resource constraints, the fully digital and self-guided structure of c-bnt makes it particularly suitable for integration into military health services. Modular designs of c-bnt allows flexible use across different phases of the deployment cycle, including pre-deployment preparation, mental health support during deployment, and post-deployment reintegration.Future studies should focus on the optimization of training duration and training tasks and identify most appropriate time points and subgroups for c-bnt. In this context, trauma chronicity, comorbidities, and prior treatment history are likely to be key factors. Moreover, studies embedded in real-world military settings are needed to test the feasibility, acceptability, and impact of c-bnt under operational conditions.