Introduction
The pathophysiology of central auditory dysfunction intersects the bounds of neurocognition, phenomenology and prognosis in schizophrenia (SZ). Disruptions in early auditory processing (EAP) give rise to altered sensory and perceptual experiences – ranging from the misrepresentation of simple sounds to complex clinical features; e.g. aberrant salience and auditory hallucinations (Iliadou et al., Reference Iliadou, Apalla, Kaprinis, Nimatoudis, Kaprinis and Iacovides2013; Javitt & Freedman, Reference Javitt and Freedman2015; Martin et al., Reference Martin, Bartolomeo, Howell, Hetrick, Bolbecker, Breier and O'Donnell2018; McKay, Headlam, & Copolov, Reference McKay, Headlam and Copolov2000; Wexler, Stevens, Bowers, Sernyak, & Goldman-Rakic, Reference Wexler, Stevens, Bowers, Sernyak and Goldman-Rakic1998). Hence, EAP abnormalities have pervasive consequences on neurocognition, psychosocial functioning and quality of life for individuals with SZ and other central nervous system disorders (Brown & Kuperberg, Reference Brown and Kuperberg2015; Revheim et al., Reference Revheim, Corcoran, Dias, Hellmann, Martinez, Butler and Javitt2014; Thomas et al., Reference Thomas, Green, Hellemann, Sugar, Tarasenko, Calkins and Light2017).
Given the prevalence of EAP abnormalities in SZ spectrum disorders, interventions targeting auditory system plasticity are being investigated as prospective therapies for cognitive and psychiatric rehabilitation (Dondé et al., Reference Dondé, Luck, Grot, Leitman, Brunelin and Haesebaert2017; Fisher, Holland, Merzenich, & Vinogradov, Reference Fisher, Holland, Merzenich and Vinogradov2009; Molina et al., Reference Molina, Joshi, Nungaray, Thomas, Sprock, Clayson and Light2021; Näätänen & Kähkönen, Reference Näätänen and Kähkönen2009; Swerdlow, Bhakta, & Light, Reference Swerdlow, Bhakta and Light2018). Auditory-based targeted cognitive training (TCT) is a ‘bottom-up’ intervention that leverages early sensory and perceptual learning mechanisms to stimulate neuroplasticity in brain networks that mediate higher-order neurocognition. TCT and other forms of cognitive training can produce meaningful gains in neurocognition and enhance measures of everyday functioning and life quality in individuals in SZ and other neuropsychiatric disorders (Prikken, Konings, Lei, Begemann, & Sommer, Reference Prikken, Konings, Lei, Begemann and Sommer2019; Tseng, DuBois, Biagianti, Brumley, & Jacob, Reference Tseng, DuBois, Biagianti, Brumley and Jacob2023; Weihing, Chermak, & Musiek, Reference Weihing, Chermak and Musiek2015; Wykes, Huddy, Cellard, McGurk, & Czobor, Reference Wykes, Huddy, Cellard, McGurk and Czobor2011). However, modest effect sizes and high inter-individual variability have held back the widespread adoption of TCT in clinical practice.
Clinical and translational studies of early auditory information processing have identified promising behavioral and neurophysiologic biomarkers linked to cognitive and functional outcomes in SZ (Joshi et al., Reference Joshi, Molina, Braff, Green, Gur, Gur and Light2023b; Light et al., Reference Light, Joshi, Molina, Bhakta, Nungaray, Cardoso and Swerdlow2020; Light & Swerdlow, Reference Light and Swerdlow2015; Thomas et al., Reference Thomas, Green, Hellemann, Sugar, Tarasenko, Calkins and Light2017). Despite this progress, clinically actionable strategies for personalized cognitive rehabilitation in SZ are lacking. Auditory processing speed (APS) – a psychophysical index of auditory processing efficiency – is the most widely studied behavioral measure of target engagement and early therapeutic sensitivity to TCT. Higher APS thresholds (indicating slower processing and greater inefficiency) are commonly found in SZ patients relative to healthy subjects and are associated with greater neurocognitive impairment (Ramsay et al., Reference Ramsay, Schallmo, Biagianti, Fisher, Vinogradov and Sponheim2020; Swerdlow et al., Reference Swerdlow, Tarasenko, Bhakta, Talledo, Alvarez, Hughes and Light2017; Tarasenko et al., Reference Tarasenko, Perez, Pianka, Vinogradov, Braff, Swerdlow and Light2016).
Our group has taken an experimental medicine approach toward understanding the role of EAP dysfunction in TCT-induced learning and brain plasticity in SZ. We have reported that robust changes in APS efficiency (i.e. learning) occur within the first hour of training (Joshi et al., Reference Joshi, Gonzalez, Molina, MacDonald, Min Din, Minhas and Light2023a; Perez et al., Reference Perez, Tarasenko, Miyakoshi, Pianka, Makeig, Braff and Light2017; Tarasenko et al., Reference Tarasenko, Perez, Pianka, Vinogradov, Braff, Swerdlow and Light2016). APS gains were paralleled by changes in auditory-evoked potentials and shifts in the spatiotemporal dynamics of distributed cortical systems during early auditory information processing (Koshiyama et al., Reference Koshiyama, Miyakoshi, Thomas, Joshi, Molina, Tanaka-Koshiyama and Light2020; Perez, Miyakoshi, Makeig, & Light, Reference Perez, Miyakoshi, Makeig and Light2019). Furthermore, we have demonstrated that training-induced gains in APS efficiency can be augmented by specific pro-cognitive drugs underscoring the potential for the synergistic enhancement of central auditory system plasticity by pharmacologic interventions (Swerdlow et al., Reference Swerdlow, Tarasenko, Bhakta, Talledo, Alvarez, Hughes and Light2017, Reference Swerdlow, Bhakta, Talledo, Kotz, Roberts, Clifford and Light2020).
The roles of APS and EAP dysfunction in predicting longitudinal outcomes in response to TCT are incompletely understood and few studies have examined APS learning dynamics over the course of training. Whereas APS is expected to improve as a function of auditory training, these gains are thought to plateau after ~20-hours (Biagianti, Fisher, Neilands, Loewy, & Vinogradov, Reference Biagianti, Fisher, Neilands, Loewy and Vinogradov2016; Fisher et al., Reference Fisher, Loewy, Carter, Lee, Ragland, Niendam and Vinogradov2015; Keefe et al., Reference Keefe, Vinogradov, Medalia, Buckley, Caroff, D'Souza and Stroup2012). Interestingly, changes in the shape of individual APS learning curves (e.g. the change from baseline after 20 h of training) were associated with the degree of neurocognitive improvement in response to TCT (Biagianti et al., Reference Biagianti, Fisher, Neilands, Loewy and Vinogradov2016). Given the rehabilitative capacity of auditory system plasticity and the heterogeneity of therapeutic responses to TCT, a detailed understanding of the timeline of TCT learning at earlier stages of training is necessary to enhance the clinical efficacy and translatability of this intervention.
The TCT platform builds on the perceptual learning operations involved in APS and embeds them within increasingly complex tasks designed to strengthen core verbal learning, working memory, and language operations. No study to date has examined the longitudinal trajectories of TCT learning metrics across the broader training platform. Here, we present a comprehensive data-driven analysis of longitudinal behavioral performance from a widely used TCT program collected as part of a randomized controlled trial evaluating the effects of TCT on neurocognition in patients with treatment-refractory SZ. Our goals were to (1) clarify the time course of APS learning trajectories, (2) evaluate the trajectories of behavioral performance from exercises in the broader TCT suite, and (3) determine whether APS and other TCT-derived learning metrics were associated with neurocognitive outcomes after a 30-hour course of training.
Methods
Participants and study design
Forty participants, including patients with treatment-refractory psychotic disorders (n = 21) and healthy subjects (HS; n = 19) were recruited from the community. Patients with chronic psychotic disorders were recruited from a community-based long-term care facility and were under public conservatorship. Diagnosis was confirmed using an abbreviated version of the Structured Clinical Interview for DSM-IV-TR (First, Spitzer, Gibbon, & Williams, Reference First, Spitzer, Gibbon and Williams2002). Exclusion criteria included a history of significant neurological illness, head injury, or hearing loss; premorbid intellectual disability; limited English proficiency; inability to provide informed consent; or a positive screen on urine toxicology.
After initial screening, participants underwent comprehensive clinical and cognitive testing and completed 1-hour (h) of auditory-based TCT delivered on individual laptops with headphones. SZ patients additionally completed up to 30-h of TCT and repeated structured clinical and cognitive assessments approximately 10–12 weeks later. Primary clinical and cognitive outcomes of this randomized controlled trial were previously reported (Hochberger et al., Reference Hochberger, Joshi, Thomas, Zhang, Bismark, Treichler and Light2019; Molina et al., Reference Molina, Thomas, Joshi, Hochberger, Koshiyama, Nungaray and Light2020a; Thomas et al., Reference Thomas, Bismark, Joshi, Tarasenko, Treichler, Hochberger and Light2018). All experimental procedures were approved by the Institutional Review Board at the University of California, San Diego (IRB#130874) and complied with the ethical standards of the Helsinki Declaration of 1975, as revised in 2008.
Cognitive and clinical assessments
Neurocognitive outcomes were assessed with the MATRICS Consensus Cognitive Battery (MCCB), which evaluates cognitive domains relevant to SZ and was designed for use as a repeated measure in clinical trials of pro-cognitive therapeutics (Nuechterlein et al., Reference Nuechterlein, Green, Kern, Baade, Barch, Cohen and Marder2008). MCCB yields age and gender corrected T-scores for individual cognitive domains (e.g. speed of processing, attention/vigilance, working memory, verbal learning, visual learning, and reasoning and problem solving) and a neurocognitive composite score. Clinical symptoms in SZ patients were assessed with the Scale for the Assessment of Positive Symptoms (SAPS) and the Scale for the Assessment of Negative Symptoms (SANS) (Andreasen, Reference Andreasen1983, Reference Andreasen1984)).
Targeted cognitive training
TCT consists of six computerized exercises (BrainHQ, Posit Science Corporation; San Francisco, CA). Collectively, these exercises target early sensory mechanisms that support auditory perception and processing speed (Sound Sweeps, Fine Tuning) and auditory working memory (Syllable Stacks, Memory Grid, To-Do List, Rhythm Recall). Training was structured into blocks that delivered carefully designed stimulus sets; thresholds were determined by varying temporal and psychophysical parameters of the stimulus sets to allow for continuous learning and improvement at 80% criterion accuracy. Descriptions of the TCT exercises and exercise-specific thresholds are provided below:
Sounds Sweeps: a time-order-judgement task where participants identify the direction of change (i.e. up v. down) of two sequential frequency-modulated sweeps. APS thresholds (in milliseconds (ms)) were estimated on a trial-by-trial basis by modulating the ratio of the interstimulus interval (ISI) and tone duration by 0.125 logarithmically-spaced step sizes using an adaptive n-up/m-down algorithm. Fine Tuning: a syllable discrimination task where participants listen to a syllable and indicate which one was presented from a pair of two confusable syllables (e.g. da v. ga) as phonological and temporal features of the stimuli were varied; thresholds were defined by the maximum level of similarity to accurately discriminate between syllable pairs; higher scores reflect better performance. Syllable Stacks: a serial memory span task where participants were presented with a series of syllables and were then asked to repeat them in order; thresholds were determined by the number of syllables recalled. Memory Grid: a visuospatial working memory task where serially presented cards representing syllables were matched; thresholds were determined by the number of syllable pairs correctly matched. To-Do List: a verbal sequence learning task where items from a list of spoken instructions were selected from a grid of everyday items (e.g. plant, bucket, shovel); thresholds were determined the number of instructions recalled. Rhythm Recall: a rhythm memory span task where participants recreate the timing and duration of brief musical melodies/rhythmic sequences; thresholds were determined by the number of beats recalled. Thresholds in Syllable Stacks, Memory Grid, To-Do List Training, and Rhythm Recall were determined by the set size (i.e. length of the memory set) where higher scores indicate better performance. Further details regarding the TCT exercises and stimulus parameters can be found in the Supplemental Materials.
Modeling learning trajectories and statistical analyses
Psychophysical thresholds were extracted from subject-specific performance files for each exercise using custom scripts written in Python (v3.7). The algorithm partitioned behavioral performance into temporally defined groupings (i.e. 1, 5, 10, 20, and 30 h) based on the recorded date and timestamp of each completed block and/or level for each exercise, separately. Local minima or maxima were extracted from all exercises at each time bin to account for exercise-specific changes in psychophysical efficiency over the course of training. Specifically, local minima were extracted from the Sound Sweeps exercise (as lower thresholds indicate faster/better APS performance), and local maxima were extracted for all other training exercises. Note: Baseline APS thresholds were operationally defined as the initial performance gains in processing speed observed during the first level (i.e. the first 20 trials) of Sound Sweeps training and were indexed as time point 0 (Figs 1a and 2b).
Figure 1. TCT behavioral performance as a function of time and exercise in schizophrenia patients. Behavioral thresholds are on the y-axis and time (hours) is on the x-axis. For perspective, the upper and lower bounds of exercise specific performance thresholds are provided in the following general form: threshold; range [min, max]. (a) Sound Sweeps: APS thresholds were defined in milliseconds (ms; range: [16ms–1000ms]). (b) Fine Tuning: thresholds were defined in terms of speech similarity (similarity index; range: [1–15]). (c) Syllable Stacks: thresholds were determined by the set size (number of syllables recalled: range: [1–12]). (d) Memory Grid: thresholds were determined by the set size (number of syllable pairs matched; threshold range: [1–10]). (e) To-Do List: thresholds were determined by the set size (number of instructions recalled; threshold range: [1–9]). (f) Rhythm Recall: thresholds were determined by the set size (number of beats recalled; threshold range: [1–8]). Note: In Sound Sweeps lower scores reflect better (i.e. faster) performance; in all other exercises, higher scores reflect better performance. Data in all subplots were presented as means ± standard deviations.
Figure 2. EAP status has differential effects on APS learning and its longitudinal trajectory. (a) Robust gains in APS efficiency were made during the first hour of training and training-driven gains were maintained throughout training. (b) EAPimpaired schizophrenia patients learned more during 1-h of TCT relative to both unimpaired schizophrenia patients (EAPwnl; d = 1.2, p = 0.002) and healthy subjects (HS; d = 1.4, p = 0.004). (c) Baseline APS was associated with general neurocognitive function in our cohort (r = 0.52, p < 0.001). Note: EAP status was determined based off the median split of baseline APS measurements from the entire sample.
This mode of temporal segmentation for psychophysical thresholds was selected a priori to (1) parallel our experimental medicine design (all participants underwent 1-h of Sound Sweeps training), (2) ensure adequate sampling of psychophysical thresholds from all training exercises (as the timing, order, and number of completed blocks/levels for each exercise varied as a function of individual subject performance and were determined by the TCT training algorithm), and (3) evaluate whether exercise-specific psychophysical thresholds and/or learning trajectories predicted therapeutic outcomes.
Three measures of target engagement were collected for each exercise: (1) baseline performance – the initial psychophysical threshold established at the first exposure for any given exercise; (2) number of blocks – a proxy for overall training exposure from each exercise; (3) peak performance – the ‘best’ psychophysical threshold reached at any point over the course of training.
Group differences on clinical and demographic variables were evaluated with independent sample t tests. Linear mixed-effects models were used to evaluate the effects of time – and where appropriate, group – on psychophysical thresholds; exercises were modeled separately with random intercepts for subjects. F-statistics were calculated as complementary measures of overall model significance for each exercise. Where appropriate, significant main effects and/or interactions were followed up with post hoc pairwise testing reported as effect sizes (Cohen's d). Exploratory correlational analyses were conducted to examine associations between the exercise-specific measures of target engagement and neurocognitive outcomes. MCCB change scores (i.e. the difference between post- and pre-training T-scores on individual MCCB domains) were the primary neurocognitive outcomes evaluated in this study. Statistical analyses were implemented using the ‘lme4’ (Bates, Mächler, Bolker, & Walker, Reference Bates, Mächler, Bolker and Walker2015) and ‘EMAtools’ packages and built-in R functions.
Results
Demographics and clinical features
Demographics and clinical characteristics are summarized in Table 1. Compared to age- and sex-matched controls, SZ patients completed fewer years of formal education (d = 1.3, p < 0.001), had lower premorbid IQ as measured by the Wide Range Achievement Test (WRAT) (d = 1.3, p < 0.001), and were significantly impaired on all cognitive domains tested (d ≈ 0.8–2.5).
Table 1. Demographic, clinical and cognitive characteristics of the sample
Modeling training-related learning effects in SZ
Figure 1 shows TCT learning trajectories as a function of time and training exercise. A summary of the number of blocks and levels completed in each exercise is available in online Supplementary Table 1. There were significant training effects on Sound Sweeps performance (F = 17.3, p < 0.0001), which were driven by robust learning effects after 1-h of training (Fig. 1a; β = 42.7, s.e. = 4.6, p < 0.0001) that persisted over the course of training. Similarly, training-related effects were noted in the auditory discrimination exercise, Fine Tuning (F = 13.3, p < 0.0001), and the auditory working memory exercise, To-Do List (F = 41.4, p < 0.0001). Main effects of time were also seen in Syllable Stacks (F = 5.8, p = 0.002), and Rhythm Recall (F = 4.8, p = 0.005). No significant main effects of time were noted in the Memory Grid exercise (F = 1.3, p = 0.28). Associations between putative measures of target engagement (e.g. baseline and best performance) and neurocognitive outcomes in SZ patients are summarized in online Supplementary Fig. 1 and online Supplementary Fig. 2.
APS is abnormal in SZ and impacts learning trajectories and sensitivity to TCT
By design, all subjects underwent 1-h of Sounds Sweeps training to identify measures of early treatment sensitivity. Relative to HS, SZ patients had higher APS thresholds at baseline (d = 0.96, p < 0.005), reflecting the fact that they required longer ISI's to accurately discriminate sweep stimuli. Linear mixed effects models revealed a significant main effect of diagnosis (β = 67.3, s.e. = 20.9, p = 0.003), time (β = −82.6, s.e. = 15.3, p < 0.0001), and a diagnosis × time interaction (Fig. 2b; β = −40.3, s.e. = 15.3, p = 0.012) on APS thresholds after 1-h of training. Consistent with prior reports (Ramsay et al., Reference Ramsay, Schallmo, Biagianti, Fisher, Vinogradov and Sponheim2020; Swerdlow et al., Reference Swerdlow, Tarasenko, Bhakta, Talledo, Alvarez, Hughes and Light2017; Tarasenko et al., Reference Tarasenko, Perez, Pianka, Vinogradov, Braff, Swerdlow and Light2016), baseline APS was strongly correlated with general neurocognitive ability in the overall sample (Fig. 2c; r = 0.52, p < 0.001).
Previous studies have suggested that the presence of EAP dysfunction may help identify subgroups of SZ patients with distinct neurocognitive profiles and sensitivities to specific pro-cognitive interventions (Dondé et al., Reference Dondé, Martínez, Kantrowitz, Silipo, Dias, Patel and Javitt2019; Medalia, Saperstein, Qian, & Javitt, Reference Medalia, Saperstein, Qian and Javitt2019; Molina et al., Reference Molina, Joshi, Nungaray, Thomas, Sprock, Clayson and Light2021). Therefore, we investigated whether EAP dysfunction influenced TCT learning, APS trajectories, and neurocognition. SZ patients were considered ‘impaired’ (EAPimpaired) if their baseline APS thresholds were above the sample median of 149 ms. Follow up linear mixed effects modeling revealed that 1-h learning effects were most pronounced in EAPimpaired SZ patients (Fig. 2b; β = −126.0, s.e. = 29.9, p < 0.001), whereas ‘unimpaired’ patients (EAPwnl; i.e. ‘within normal limits’) performed similar to HS in terms of baseline APS (Fig. 2a, d = 0.25) and 1-h learning (Fig. 2b, d = 0.30). Interestingly, EAPimpaired patients required significantly longer training periods to reach their ‘peak’ APS performance (t = 3.38, d = 1.13, p < 0.005); on average (mean ± s.d.) EAPimpaired patients required 12.1 ± 10.8 h to reach their peak psychophysical performance v. 1.9 ± 2.2 h of training in EAPwnl patients.
Given the heterogeneity of neurocognitive responses to TCT, we asked whether baseline EAP dysfunction was predictive of gains in verbal learning – a key rehabilitative target of TCT and a mediator of functional outcomes in SZ. Indeed, verbal learning gains (i.e. the post- minus pre-training difference score from the MCCB Verbal Learning domain) were most pronounced in EAPimpaired patients (Fig. 3a; t = 2.98, d = 1.45, p = 0.013). Receiver operating characteristic (ROC) curves corroborate this finding and reveal similar sensitivity profiles between baseline APS and 1-h learning (Fig. 3b).
Figure 3. Baseline APS predicts verbal learning gains after 30-h of TCT in schizophrenia patients. (a) EAPimpaired schizophrenia patients had greater verbal learning gains relative to ‘unimpaired’ patients (d = 1.4, p = 0.01). (b) Receiver operating characteristic (R.O.C) analysis of baseline and 1-h APS learning suggest that APS is a sensitive predictor of future verbal learning gains in schizophrenia patients. Note: APST0 refers to baseline APS.
Discussion
Despite the general consensus view that cognitive remediation is beneficial in the treatment of chronic psychotic disorders (Keepers et al., Reference Keepers, Fochtmann, Anzia, Benjamin, Lyness and Mojtabai2020), TCT and other forms of cognitive rehabilitation remain underutilized in real-world clinical settings. The field has made substantial progress toward understanding the biological and behavioral mechanisms of cognitive training and learning-induced neuroplasticity in SZ (Biagianti, Bigoni, Maggioni, & Brambilla, Reference Biagianti, Bigoni, Maggioni and Brambilla2022; Dale et al., Reference Dale, Brown, Fisher, Herman, Dowling, Hinkley and Vinogradov2016; Koshiyama et al., Reference Koshiyama, Miyakoshi, Thomas, Joshi, Molina, Tanaka-Koshiyama and Light2020; Molina et al., Reference Molina, Thomas, Joshi, Hochberger, Koshiyama, Nungaray and Light2020a), yet clinically actionable tools to stratify SZ patients to TCT v. an alternative behavioral and/or pharmacologic intervention based on their likelihood of therapeutic response are lacking. Studies looking ‘under the hood’ of behavioral performance in TCT may help identify possible substrates underlying TCT sensitivity that may facilitate future personalized medicine approaches for cognitive rehabilitation.
This study confirms the presence of APS abnormalities in SZ and provides further evidence of the relationships between APS, neurocognition, and TCT-related outcomes. Moreover, we clarify the trajectory of perceptual learning as a function of training in TCT. Previous studies have reported that APS learning ‘plateaus’ after 20 h of training and have suggested that individual differences in APS learning curves may predict verbal learning gains after a 40-h course of TCT (Biagianti et al., Reference Biagianti, Fisher, Neilands, Loewy and Vinogradov2016). Our data reveal that the training-induced ‘plateau’ in APS occurs much earlier than previously thought and suggest that a substantial amount of ‘learning’ occurs within the first hour of training. Further, we found that SZ patients learned more than HS during this brief 1-h exposure to TCT, replicating previous experimental medicine reports (Perez et al., Reference Perez, Tarasenko, Miyakoshi, Pianka, Makeig, Braff and Light2017; Swerdlow et al., Reference Swerdlow, Tarasenko, Bhakta, Talledo, Alvarez, Hughes and Light2017, Reference Swerdlow, Bhakta, Talledo, Kotz, Roberts, Clifford and Light2020; Tarasenko et al., Reference Tarasenko, Perez, Pianka, Vinogradov, Braff, Swerdlow and Light2016).
Results suggest that EAP dysfunction, as indexed by baseline psychophysical efficiency (i.e. APS ⩾ 150), may indicate a greater capacity for change and/or sensitivity to auditory perceptual training given that the individuals who learned the most during the first hour were the SZ patients with the greatest EAP deficits (EAPimpaired). Interestingly, APS learning curves were markedly different in the EAPimpaired subgroup; and these individuals also benefited from greater Verbal Learning gains in response to a therapeutic course of TCT relative to SZ patients with intact EAP function (EAPwnl). These findings are consistent with the growing body of clinical and translational evidence in support of EAP dysfunction as a pathophysiological marker for a discrete subtype of SZ with distinct clinical features, illness course, and sensitivity to sensory-based therapies like TCT (Dondé et al., Reference Dondé, Martínez, Kantrowitz, Silipo, Dias, Patel and Javitt2019; Hochberger et al., Reference Hochberger, Joshi, Thomas, Zhang, Bismark, Treichler and Light2019; Kambeitz-Ilankovic et al., Reference Kambeitz-Ilankovic, Wenzel, Haas, Ruef, Antonucci, Sanfelici and Biagianti2020; Medalia et al., Reference Medalia, Saperstein, Javitt, Qian, Meyler and Styke2023; Reference Medalia, Saperstein, Qian and Javitt2019; Molina et al., Reference Molina, Joshi, Nungaray, Thomas, Sprock, Clayson and Light2021, Reference Molina, Thomas, Joshi, Hochberger, Koshiyama, Nungaray and Light2020a, Reference Molina, Voytek, Thomas, Joshi, Bhakta, Talledo and Light2020b).
These findings do not negate the requirement for additional training and practice beyond the first hour to benefit from the salutary effects of TCT. It is conceivable that as individuals continue to practice and refine their newly acquired perceptual skills over time, the increase in APS efficiency may translate into a decreased ‘cost’ on higher-order attentional and working memory systems that is needed to maintain the same level of performance; this may help subjects apply learning from the stimulus-bound sensory and perceptual conditions in TCT to gains in everyday functions like communication and the management of activities of daily living.
An alternative possibility is that the differences in learning and APS gains made by SZ patients were driven by their underlying neurocognitive deficits v. EAP dysfunction and/or their sensitivity to TCT. Though implicit learning mechanisms are generally thought to be preserved in SZ, patients with greater neurocognitive impairment may require additional time and training to learn the mechanics of the TCT exercises and arrive at the same endpoint (Gomar et al., Reference Gomar, Pomarol-Clotet, Sarró, Salvador, Myers and McKenna2011; Horan et al., Reference Horan, Green, Knowlton, Wynn, Mintz and Nuechterlein2008; Weickert et al., Reference Weickert, Terrazas, Bigelow, Malley, Hyde, Egan and Goldberg2002). To this end, we found that EAPimpaired patients required longer training intervals to reach their peak psychophysical threshold (i.e. their ‘best’ performance) relative to unimpaired SZ patients; however, there were no significant differences in terms of general neurocognition or verbal learning on pre-training MCCB scores between the EAP subgroups. Future studies conducted in larger samples should evaluate the role of individual differences in neurocognitive abilities as potential mediators and/or moderators of perceptual learning in TCT.
Another consideration is that multiple learning processes are likely occurring simultaneously during the early stages of auditory training. While this study was not explicitly designed to investigate the ‘phases’ of learning, the APS learning curves seen in SZ patients are in general agreement with the observed dynamics of experience-dependent plasticity seen in studies of perceptual learning across species and other forms of learning (Ahissar, Nahum, Nelken, & Hochstein, Reference Ahissar, Nahum, Nelken and Hochstein2009; Ahissar & Hochstein, Reference Ahissar and Hochstein1997; Dosher & Lu, Reference Dosher and Lu2007; Hong, Gallanter, Müller-Oehring, & Schulte, Reference Hong, Gallanter, Müller-Oehring and Schulte2019; Lengyel & Fiser, Reference Lengyel and Fiser2019; Palva et al., Reference Palva, Zhigalov, Hirvonen, Korhonen, Linkenkaer-Hansen and Palva2013). Curiously, statistical modeling of human psychophysical performance during auditory perceptual training suggest that early performance gains are largely driven by perceptual learning (experience-dependent changes in sensory representation) and to a lesser degree by procedural learning (learning the task) (Hawkey, Amitay, & Moore, Reference Hawkey, Amitay and Moore2004), which are corroborated by the findings of neuroplastic changes that occur at the level of sensory cortices after brief exposures of perceptual training (Alain, Snyder, He, & Reinke, Reference Alain, Snyder, He and Reinke2007; Koshiyama et al., Reference Koshiyama, Miyakoshi, Thomas, Joshi, Molina, Tanaka-Koshiyama and Light2020; Perez et al., Reference Perez, Miyakoshi, Makeig and Light2019).
Interestingly, we observed strong interrelationships between APS measures and the other TCT-derived performance metrics (online Supplementary Fig. 1). We previously reported the pharmacologic enhancement of auditory perceptual learning by specific pro-cognitive drugs. It is conceivable that the covariance structure derived from the TCT learning metrics may be leveraged to determine: (1) whether pro-cognitive drugs enhance specific phases/stages of learning, (2) the optimal entry/endpoints for specific augmentation strategies, and (3) TCT dosing schedules (i.e. how much training is needed to yield a desired clinical response). Hence, the present study suggests a possible framework to track and monitor individual progress during training and to evaluate the effects of personalized augmentation strategies (e.g. behavioral, pharmacologic, and/or neuromodulatory) on individual learning trajectories.
To our knowledge, this is first study to investigate the longitudinal trajectory of performance metrics derived from the broader TCT program. Our data-driven approach leveraged all available behavioral data from each subject and from each exercise; this allowed us to capture between- and within-subject variability in psychophysical performance at earlier timeframes and with greater temporal resolution than previously reported. Individual associations between these putative measures of target engagement and neurocognitive outcomes should be interpreted with caution given the modest sample size and the fact that this study was designed as a secondary analysis of existing data (Hochberger et al., Reference Hochberger, Joshi, Thomas, Zhang, Bismark, Treichler and Light2019; Molina et al., Reference Molina, Thomas, Joshi, Hochberger, Koshiyama, Nungaray and Light2020a; Thomas et al., Reference Thomas, Bismark, Joshi, Tarasenko, Treichler, Hochberger and Light2018). That said, this study provides proof-of-concept evidence that meaningful signals can be extracted from the wealth of behavioral data collected over the entire course of training from each subject and suggest that – in doing so – we may be able to capture some of the ‘missing’ variability in TCT-related neurocognitive outcomes.
Conclusions
Identifying early predictors of individual response to TCT or other pro-cognitive interventions may help steer patients towards specific treatments and help guide therapeutic decision-making. The availability and relative ease of obtaining these behavioral data (i.e. data can be readily downloaded from the TCT application) makes it possible to monitor TCT-related behavioral performance in ‘real-time’ and potentially to tailor rehabilitative programming to optimize individual outcomes.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291724000783.
Acknowledgements
This work was made possible by VA Office of Research & Development, Rehabilitation Research & Development Service CDA1 grant 1IK1RX003683 (PI: Molina) and Brain and Behavior Research Foundation Young Investigator Award #28927 (PI: Molina). Additional support was provided by R33-MH125114 (PI: Light), R33-MH123603 (PI: Swerdlow), R01-AG059640 (PI: Swerdlow), VA CDA2 RRD 1IK2RX003395 (PI: Joshi), and the Sidney R. Baer, Jr. Foundation. The contents of this report do not represent the views of the U.S. Department of Veterans Affairs or the United States Government. Sponsors had no influence or role in the study.
Disclosures
Dr Light is a consultant for Johnson & Johnson, Neurocrine, NeuroSig, and Sosei-Heptares. Dr Attarha is an employee at Posit Science, the company that develops BrainHQ. All other authors have no actual or potential conflicts to report, financial or otherwise.
