Neurocognitive deficits are core features of schizophrenia that typically affect several cognitive domains, with an impairment usually involving memory, attention and executive functions and a less prominent impairment in the domains of learning, vocabulary and visual perceptual skills.Reference Green1 These deficits appear to be related to poor work and school functioning, and poor social functioning outcomes.Reference Green1,Reference Sheffield, Karcher and Barch2
Neurocognitive difficulties do not belong only to overt psychotic disease or to schizophrenia, as they can be present in the preclinical phaseReference Green1,Reference Fusar-Poli, Deste, Smieskova, Barlati, Yung and Howes3,Reference Fusar-Poli, Rocchetti, Sardella, Avila, Brandizzi and Caverzasi4 in patients at clinical high risk for psychosis (CHR). When considering psychotic disorders as a spectrum ranging from the CHR phase to first-episode psychosis to chronic schizophrenia, cognitive deficits seem to characterise all phases of the illness, with increasing severity.Reference Sheffield, Karcher and Barch2,Reference Cornblatt, Obuchowski, Roberts, Pollack and Erlenmeyer-Kimling5,Reference Rapoport, Giedd and Gogtay6 Those deficits mainly involve attention and executive function, speed of processing, verbal learning,Reference Fusar-Poli, Deste, Smieskova, Barlati, Yung and Howes3,Reference Johnstone, Ebmeier, Miller, Owens and Lawrie7–Reference Lin, Yung, Nelson, Brewer, Riley and Simmons11 working memory and declarative memory, especially in the CHR phase that transitions to psychosis.Reference Seidman, Shapiro, Stone, Woodberry, Ronzio and Cornblatt12 Moreover, as in overt schizophrenia, the cognitive impairment in the CHR phase can be accompanied by a drop in functioning.Reference Fusar-Poli, Rocchetti, Sardella, Avila, Brandizzi and Caverzasi4
Neurocognitive skills and depression
Furthermore, deficiencies involving some neurocognitive skills and daily functioning may also be present in other psychiatric disorders, such as depression.Reference Sheffield, Karcher and Barch2,Reference Zanelli, Reichenberg, Morgan, Fearon, Kravariti and Dazzan13,Reference Mallawaarachchi, Amminger, Farhall, Bolt, Nelson and Yuen14 A recent study showed that neurocognitive skills were related to scores measuring work/school functioning, as well as to those measuring the ability to cope with symptoms, in a sample including patients at CHR and patients with recent-onset depression (ROD).Reference Squarcina, Kambeitz-Ilankovic, Bonivento, Prunas, Oldani and Wenzel15 A meta-analysis by Sheffield et alReference Sheffield, Karcher and Barch2 found that patients with psychotic depression show impaired IQ and cognitive impairment involving verbal learning and attention since the first episode of psychosis. The authors suggest the cognitive profile in psychotic depression is comparable to the cognitive profile in psychosis without depression. However, patients with psychotic depression are also reported to be significantly more impaired than patients with non-psychotic depression. Moreover, global cognitive impairment was reported to accompany depression, even in the absence of psychotic symptoms. Finally, general intelligence deficits are present in the CHR state with comorbid depression at 12 months follow-up, accompanied by more specific deficits in verbal fluency and verbal memory performance at baseline. Neurocognitive deficits in the psychotic spectrum and depression appear to overlap, involving general cognitive skills, attention and memory, and the presence of those common features raises a question concerning the specificity of cognitive deficits in the preclinical states of these diseases. Moreover, a broad clinical high-risk condition emerged in recent studies, introducing a transdiagnostic approach for understanding risk factors and pathogenic mechanisms. This broad clinical high-risk conditionReference McGorry, Hartmann, Spooner and Nelson16 may be characterised by a general cognitive impairment that differentiates into psychosis or depression later on. Although approximately 25% of individuals at CHR transition to psychosis,Reference Salazar De Pablo, Soardo, Cabras, Pereira, Kaur and Besana17 this is not always the outcome. Instead, depression is frequently reported as a comorbidity in CHR (i.e. major depressive disorder (MDD), dysthymia and depression not otherwise specified).Reference Addington, Farris, Liu, Cadenhead, Cannon and Cornblatt18 At the same time, patients with MDD can show psychotic symptoms.Reference Sheffield, Karcher and Barch2,Reference Zanelli, Reichenberg, Morgan, Fearon, Kravariti and Dazzan13
Moreover, there are some methodological and data analyses challenges in these studies, which may affect our interpretation of the extent to which the cognitive profiles overlap in psychosis and depression. For example, when several neurocognitive measures, derived from several tests, are reduced to the relevant information on latent domains,Reference Seidman, Shapiro, Stone, Woodberry, Ronzio and Cornblatt12 it appears that the more evident deficits in CHR mainly concern declarative memory and sustained attention.Reference Seidman, Shapiro, Stone, Woodberry, Ronzio and Cornblatt12 In contrast, MDD patients show deficits in semantic aspects of word retrieval (as measured by semantic fluency) and attention switching.Reference Sheffield, Karcher and Barch2,Reference Zanelli, Reichenberg, Morgan, Fearon, Kravariti and Dazzan13
Although the neurocognitive profiles of CHR (i.e. characterised according to standardised criteria representing a high risk of transition to psychosis) were extensively studied, to the best of our knowledge, they were never directly compared with patients at an early stage of depression, with no subclinical or overt psychotic symptoms, or to patients with overt psychotic symptoms. For this reason, it is difficult to make evidence-based considerations on similarities and differences between the cognitive profiles in psychosis and depression, and how they differentiate from the CHR.
The current study aims to characterise cognitive profiles of individuals with recent-onset psychosis (ROP), individuals with ROD and individuals at CHR, recruited at seven sites across five European countries in comparison to healthy controls and between ROP, ROD and CHR groups. The participants were assessed by a comprehensive neuropsychological test battery, encompassing the domains of speed of processing, attention and working memory, verbal learning and spatial learning. The presence and possible influence of depression in CHR and the illness duration, as well as sociodemographic variables, were taken into account.
Method
Participants
The present investigation was carried out as part of the European project ‘Personalized Prognostic Tools for Early Psychosis Management’ (PRONIA). This is a study programme that involved seven sites across five European countries: the Departments of Psychiatry at Ludwig-Maximilian University of Munich and University of Cologne, Germany; University of Basel, Switzerland; University of Milan and University of Udine, Italy; University of Birmingham, UK and University of Turku, Finland. The PRONIA project was designed with the main goal of identifying (bio)markers associated with an enhanced risk of developing psychosis, to improve early detection and prognosis. After screening and inclusion, at baseline (time point 0), in addition to the neurocognitive evaluation that is the focus of the present work, the participants underwent a clinical assessment through the use of validated observer- and self-rated tools, the acquisition of neuroimaging data and a blood withdrawal for genetic and metabolic analyses. The examination was divided into three to four sessions of 1–2 h, attempting to keep the patient burden of data collection at the minimum.
The data collected at the baseline evaluation from 116 people at CHR, 123 people with ROD, 105 people with ROP and 372 healthy controls are analysed in the present work.
Information concerning psychotropic medications; illness duration in CHR, ROD and ROP; frequencies of DSM-IV axis 1 diagnosis in CHR and ROP and scores on the Beck Depression Inventory (BDI) are reported in Table 1.
Participant characteristics
ROD, recent-onset depression; CHR, clinical high risk for psychosis; ROP, recent-onset psychosis; HSD, honestly significant difference; FWER, family-wise error rate. *P≤0.05, **P≤0.001.
The study methods, inclusion and exclusion criteria and clinical characteristics of the participants are reported in Supplementary Material available at https://doi.org/10.1192/bjp.2023.98, and are published elsewhere.Reference Koutsouleris, Kambeitz-Ilankovic, Ruhrmann, Rosen, Ruef and Dwyer19
The study protocol was registered with the German Clinical Trials Register (identifier DRKS00005042) and received ethical approval by the research ethics committees of each site. This study was performed in accordance with the Helsinki Declaration of 1975, as revised in 2008. All participants provided written informed consent for participation in the study.
Measures
The PRONIA Cognitive Battery (PCB) is a comprehensive battery of ten neuropsychological tests selected as measuring those cognitive skills that, based on previous literature,Reference Green, Nuechterlein, Kern, Baade, Fenton and Gold20,Reference Nuechterlein, Green, Kern, Baade, Barch and Cohen21 are reportedly altered in patients with psychosis or depression.
The ten tests were implemented in the software programme Psychology Experiment Building Language (SourceForge, San Diego, CA, USA; https://pebl.sourceforge.net), running on Windows 8 (or later). As PRONIA is a multicentric European project, involving centres located in Germany, Italy, Switzerland, the UK and Finland, the test's battery was implemented in four different languages.
From the original set of ten tests, eight were used in the present work. The Diagnostic Analysis of Non-Verbal Accuracy – Affective Faces TrialReference Nowicki S, Duke, Hall and Bernieri22 was excluded because of its ceiling effect across the four groups. Also, the Salience Attribution TaskReference Rosseel23 was excluded because it is an experimental task that will be better analysed in a next work. From those eight tests, 16 relevant measures were derived (12 accuracy scores, two variables on time of execution, one reaction time variable and one error variable). The list of tests is reported in Table 2. More details are reported in the Supplementary Material.
List of tests included in the PRONIA Cognitive Battery
The WAIS-4, DANVA-2-AF and SAT-SV were not included in the present work. The WAIS-4 was administered as a screening test for excluding participants whose estimated IQ was ≤70. The DANVA-2-AF had a clear ceiling effect within all the groups, whereas the SAT-SV is an experimental task that will be analysed in more depth in future work.
The present study includes the data that were acquired at the first visit (baseline) after the participants’ screening and enrolment.
Data analysis
The data analysis included two steps. At first, the raw measures, obtained by healthy controls and patients, were reduced to a hypothesis-driven factorial model by means of a confirmatory factor analysis (CFA). Then, the factor scores were derived by imposing the model that proved to represent the PCB's factorial structure most appropriately. The factor scores were finally entered into the group comparisons to compare the cognitive performances between the groups.
CFA
Preliminary to the group comparisons, 16 raw measures drawn from the eight tests were entered in a CFA, along with gender, gender and education as covariates.
The CFA was carried out with the lavaan 0.6–10 package on R version 4.1.3 (R Foundation for Statistical Computing, Vienna, Austria; https://cran.r-project.org/bin/windows/base/old/4.1.3/).Reference Seabold and Perktold24 Data were fitted to five-, four- and three-factor a priori models. Then, factors were reduced to four and three, merging those factors that appeared to share common sources of variance at each step. All of the models were tested with the maximum likelihood method. For each model the χ 2-test evaluated the difference between the observed and expected covariance matrices, with values closer to 0 indicating a better fit. The Akaike information criterion (AIC) and Bayesian information criterion (BIC) were used to compare the models with five, four and three factors. Between each pair, the model that minimized those indexes was considered the preferred model. The procedures used for the CFA and the indexes that guided the factorial model selection are reported in the Supplementary Material.
Age, site, gender, years of education and language were used as covariates in the CFA, to factor out the confounding effect of those features.
Group comparisons
Upon selection of the appropriate factorial model that best represented the constructs measured by the PCB, the factor scores were extracted by fitting each participants’ raw scores to the four-factor model.
Each set of factor scores (one for each factor in the model) were entered as dependent variables into four univariate analyses of variance (ANOVA), each one testing for the overall effect of the group (CHR, ROD, ROP and healthy controls). If the overall model was significant, post hoc between-groups pairwise differences were tested by means of Tukey's honestly significant difference (HSD) test, which allows for control of the family-wise error rate (FWER).
Between-groups differences concerning age, years of education, estimated IQ and BDI scoreReference Beck, Steer and Brown25 were assessed by means of univariate ANOVAs. Finally, the Pearson's χ 2-test was used to see if group differences in the distribution of men and women existed.
The alpha level was set at 0.05. Note that Tukey's HSD test already controls for FWER, thus the alpha level does not need further adjustments for multiple comparisons.
As the factor scores were already corrected by age, site, gender, education, language and testing site, no covariates were needed to be entered in these analyses.
The group comparisons were performed with the statsmodels 0.8.0 Python 3.6. module (statsmodels (Perktold, Seabold, Taylor); https://www.statsmodels.org/0.8.0/).Reference Seabold and Perktold24
Results
CFA
The four-factor model including speed of processing, attention and working memory, verbal learning and spatial learning was selected as it provided the better trade-off between good fit and simplicity compared with the five- and three-factor models. Indeed, the four-factor model was comparable to (χ 2-test of difference: P = 0.0985), although simpler than (AIC: 25 350 v. 25 357; BIC: 25 645 v. 25 670), the five-factor model. Also, the three-factor model had a fit that was neither equal to (χ 2-test of difference: P < 0.001) nor simpler than (AIC: 25 370 v. 25 350; BIC: 25 651 v. 25 645) the four-factor model.
The 16 PCB measures, grouped according to their belonging to the appropriate factor, are listed in Table 3.
Tests grouped by factors
Note, see Supplementary Materials for a more detailed description of the measures derived from the PRONIA Cognitive Battery and used in this analysis. RAVLT, Rey Auditory Verbal Learning Test; ROCF, Rey-Osterrieth Complex Figure.
More details concerning the results of the CFA with five, four and three factors are reported in the Supplementary Material, along with the relevant indexes.
Group comparisons
Demographic and clinical characteristics
The sample sociodemographic, cognitive and clinical characteristics are detailed in Table 1. Overall, there was a significant main between-group difference in age (F(3, 712) = 6.44, P < 0.001). The post hoc Tukey's HSD test showed that individuals at CHR (mean 22.89 years, s.d. = 4.80) were younger than those with ROD (mean 25.51 years, s.d. = 6.04) (P-FWER < 0.05, 95% CI 0.621–4.6254) and those with ROP (mean 25.28 years, s.d. = 5.06), but they did not differ significantly from healthy controls. The ROD, ROP and healthy control groups did not differ in age (Table 1). Also, there was an overall between-group difference in the distribution of men and women (χ 2(3) = 24.41, P < 0.001). The healthy control group had a lower proportion of men (38%) and a higher proportion of women (62%) compared with the CHR group (52% men, 48% women) (χ 2(1) = 5.85, P = 0.02), as well as a higher proportion of men and a lower proportion of women compared with the ROP group (66% men, 34% women) (χ 2(1) = 3.84, P < 0.001). The proportion of men was higher in the ROP group compared with the CHR (χ 2(1) = 3.88, P = 0.05) and ROD (χ 2(1) = 7.17, P = 0.02) groups (Table 1). There was a main between-group difference in the estimated IQ (F(3,712) = 20.40, P < 0.001). Healthy controls (mean 113.50 IQ; s.d. = ±14.84) had an higher IQ than those at CHR (mean 108.28 IQ; s.d. = ±15.27) (P-FWER < 0.05, 95% CI −1.04 to −9.41), those with ROP (mean 100.48 IQ; s.d. = ±16.34) (P-FWER < 0.05, 95% CI −17.38 to −8.68) and those with ROD (mean 108.90 IQ, s.d. = ±15.69) (P-FWER < 0.05, 95% CI −8.70 to −0.51). Also, there was a significant difference in IQ between ROP and CHR groups (P-FWER < 0.05, 95% CI −13.10 to −2.50) and between ROP and ROD groups (P-FWER < 0.05, 95% CI −13.65 to −3.20). CHR and ROD groups did not differ from each other for the estimated IQ (P-FWER > 0.05). A main between-group difference also concerned the years of education (F(3, 712) = 12.95, P < 0.001). Healthy controls (mean 15.68 years, s.d. = ±3.28) had more years of educations than those at CHR (mean 13.70 years, s.d. = ±2.93) (P-FWER < 0.05, 95% CI −1.11 to −2.85) and those with ROP (mean 14.46 years, s.d. = ±3.21) (P-FWER < 0.05, 95% CI −2.13 to −0.32). Also, invidividuals with ROD (mean 14.99 years, s.d. = ±3.11) had more years of education than individuals at CHR (P-FWER < 0.05, 95% CI −1.04 to −9.41) and those with ROP (P-FWER < 0.05, 95% CI −13.65 to −3.20).
There were no differences between CHR, ROP and ROD groups in terms of illness duration (in days) (F(2 318) = 0.51, P = 0.599).
Finally, 56 individuals at CHR also had a coexisting DSM-IV axis 1 diagnosis of MDD. The duration (in days) of the axis 1 diagnosis in the CHR group (mean 247.29, s.d. = 416.74) was not significantly longer than the duration of the diagnosis in the ROD group (mean 222.86, s.d. = 213.11) (F(1, 220) = 0.33, P = 0.566). To exclude a possible effect of depressive symptoms (see Table 1) on neurocognitive performance, we ran a supplementary regression analysis that investigated the relationship between depressive symptoms, as measured by BDI score, and the neurocognitive factor scores. None of the four regression models showed either a significant main effect of the BDI scores or significant interactions between BDI score and groups (all P > 0.1) on the neuropsychological performance.
Finally, 60 out of 716 (8.38%) participants (11 at CHR, 12 with ROD, 13 with ROP, 24 healthy controls) reported the use of cannabis in the 3 months before the baseline evaluation. at an average frequency of 4.23 times a month (s.d. = ±1.5). The average frequencies of use within each group and in the whole sample are reported in Table 1. There was a significant difference between groups in the average frequency of cannabis use, with CHR and ROD groups reporting higher frequency than healthy controls and the ROP group. Also the ROP group reported higher frequency than healthy controls, whereas the CHR and ROD groups did not differ from each other (details are reported in Table 1). However, the frequency did not correlate with the neurocognitive factors (all P > 0.05).
Information concerning the proportion of pharmacological treatments between groups are also reported in Table 1.
Neurocognitive domains
The score means and s.d. derived from the neurocognitive measures in the four factorial domains (speed of processing, attention and working memory, verbal learning and spatial learning) in the four groups are reported in Supplementary Table 2.
Speed of processing
The analysis showed a significant main effect of the clinical group on this domain (F(3, 713) = 26.25, P < 0.001), with patients generally reporting worse performances than healthy controls.
The post hoc comparisons revealed that the ROP group had significantly worse performance than all other groups: healthy controls versus ROP (P-FWER < 0.05, 95% CI −1.0154 to −0.5466), ROD versus ROP (P-FWER < 0.05, 95% CI −0.9679 to −0.4043) and CHR versus ROP (P-FWER < 0.05, 95% CI −0.7402 to −0.1688). In turn, the CHR group showed impaired performance relative to healthy controls (P-FWER < 0.05, 95% CI −0.1009 to −0.552), but not to the ROD group (P-FWER > 0.05). Finally, scores for those with ROD and healthy controls did not differ significantly (P-FWER > 0.05) (see also Supplementary Fig. 1(a), Table 2).
Attention and working memory
The one-way ANOVA showed a significant main effect of the clinical group (F(3, 713) = 19.39, P < 0.001).
The ROP group performed worse than healthy controls (P-FWER < 0.05, 95% CI −0.8877 to −0.4303), the ROD group (P-FWER < 0.05, 95% CI −0.8995 to −0.3496) and the CHR group (P-FWER < 0.05, 95% CI −0.7274 to −0.1698). No other post hoc differences emerged between the other groups (all P-FWER > 0.05) (see also Supplementary Fig. 1(b), Table 2).
Spatial learning
For speed of processing and attention and working memory, patients generally had significantly poorer spatial learning scores than healthy controls (F(3, 713) = 23.39, P < 0.001). Spatial learning scores for the ROP group were significantly worse compared with all the other groups: healthy controls versus ROP (P-FWER < 0.05, 95% CI −0.8963 to −0.4636), ROD versus ROP (P-FWER < 0.05, 95% CI −0.8661 to −0.3458) and CHR versus ROP (P-FWER < 0.05, 95% CI −0.6628 to −0.1353). Moreover, scores for the CHR group were significantly below those for healthy controls (P-FWER < 0.05, 95% CI −0.0727 to −0.489), but performance for the CHR and ROD groups was comparable (P-FWER > 0.05). The ROD group did not differ significantly from healthy controls (P-FWER > 0.05) (see also Supplementary Fig. 1(c), Table 2).
Verbal learning
In the verbal learning domain, as in the others, a main effect of the clinical group was found, with patients generally performing significantly worse than healthy controls (F(3, 713) = 12.58, P < 0.001).
The ROP group performed significantly worse than healthy controls (P-FWER < 0.05, 95% CI −0.8399 to −0.3358), the ROD group (P-FWER < 0.05, 95% CI −0.8068 to −0.2007) and the CHR group (P-FWER < 0.05, 95% CI −0.6707 to −0.0562), which, in turn, did not differ from each other (all P-FWER > 0.05) (see also Supplementary Fig. 1(d), Table 2).
Discussion
The present work investigated cognitive alterations in patients across the psychotic spectrum and with ROD. To the best of our knowledge, this is the first study with such a comprehensive design, aiming to directly compare patients with depression and patients in the psychosis spectrum in early illness stage while including healthy controls.
A preliminary CFA identified four main latent factors corresponding to four cognitive domains: speed of processing, attention and working memory, verbal learning and spatial learning. Our findings show that there was an overall main effect of the clinical group on cognitive performance, with patients generally performing worse than healthy controls. In particular, the CHR group demonstrated impairment compared with healthy controls in the speed of processing and spatial learning domains, but did not differ significantly from healthy controls in the attention and working memory and verbal learning domains. The ROP group demonstrated impairment with respect to all of the other groups in all four domains. There were no significant differences in the performance of the ROD group compared with healthy controls.
In previous studies, neurocognitive deficits were reported in both psychotic spectrumReference Green1–Reference Fusar-Poli, Deste, Smieskova, Barlati, Yung and Howes3,Reference Cornblatt, Obuchowski, Roberts, Pollack and Erlenmeyer-Kimling5,Reference Johnstone, Ebmeier, Miller, Owens and Lawrie7–Reference Lin, Yung, Nelson, Brewer, Riley and Simmons11 and affective disorders.Reference Sheffield, Karcher and Barch2,Reference Zanelli, Reichenberg, Morgan, Fearon, Kravariti and Dazzan13–Reference McGorry, Hartmann, Spooner and Nelson16,Reference Allott, Fisher, Amminger, Goodall and Hetrick26 However, cognitive performance of patients with depression and patients at CHR were never directly compared. Also, neurocognitive deficits were mainly reported in patients with depression who also had psychotic symptomsReference Sheffield, Karcher and Barch2,Reference Zanelli, Reichenberg, Morgan, Fearon, Kravariti and Dazzan13 or in patients at CHR with depression as a comorbidity,Reference Mallawaarachchi, Amminger, Farhall, Bolt, Nelson and Yuen14 making it difficult to draw reliable conclusions concerning differences and similarities of these two groups. The majority of previous studies used individual test scores,Reference Fusar-Poli, Deste, Smieskova, Barlati, Yung and Howes3 which makes it difficult to understand the main latent skill that is altered in one group of patients or in another. For this reason, preliminary to the group comparisons, we performed a CFA that indicated a four-factor model as the best fit for the cognitive data derived from the neurocognitive tests administered in this study: speed of processing, attention and working memory, verbal learning and spatial learning.
The results obtained in our CHR group partly fit with the findings of previous studies. For example, deficits in speed of processingReference Seidman, Giuliano, Meyer, Addington, Cadenhead and Cannon9,Reference Anda, Brønnick, Johannessen, Joa, Kroken and Johnsen27,Reference Cannon, Yu, Addington, Bearden, Cadenhead and Cornblatt28 and spatial learningReference Seidman, Shapiro, Stone, Woodberry, Ronzio and Cornblatt12 have already been reported. However, the lack of difference between CHR and healthy controls in verbal learning and attention and working memory is not in line with other previous results.Reference Seidman, Giuliano, Meyer, Addington, Cadenhead and Cannon9,Reference Anda, Brønnick, Johannessen, Joa, Kroken and Johnsen27,Reference Cannon, Yu, Addington, Bearden, Cadenhead and Cornblatt28 Specific impairments, including speed of processing and verbal memory, can be predictive of transition to psychosis, and are used in the recently validated North American Prodrome Longitudinal Study 2 (NAPLS-2) calculator.Reference Koutsouleris, Dwyer, Degenhardt, Maj, Urquijo-Castro and Sanfelici29,Reference Van Os and Guloksuz30 Conversely, individuals at CHR in the current PRONIA study showed no deficits in verbal memory, although they appear comparable to the ROP group with respect to speed of processing. The difference between the results obtained by NAPLS-2 and PRONIA in the domain of verbal learning may be attributable to discrepancy in the assessment. The PRONIA study used the Rey Auditory Verbal Learning Test (RAVLT), whereas NAPLS-2 used the Hopkins Verbal Learning Test (HVLT). Although the HVLT includes three immediate repetition trials, which may correspond to the memory acquisition phase, the RAVLT has five repetition trials, which may also relate to the retrieval of already established memories.Reference Cambridge, Knight, Mills and Baune31 Further differences in findings of the two big cohorts, PRONIA and NAPLS-2, could be attributable to cultural or sociodemographic differences of the participants and broader health and social welfare system discrepancies.
Different from previous reports of cognitive deficits in MDD patients,Reference Sheffield, Karcher and Barch2,Reference Zanelli, Reichenberg, Morgan, Fearon, Kravariti and Dazzan13–Reference McGorry, Hartmann, Spooner and Nelson16,Reference Allott, Fisher, Amminger, Goodall and Hetrick26 in the present study the ROD group performed similarly to healthy controls in all domains. However, patients with ROD who participated in the present study were at the early stage of the disease, and cognitive deficit may not yet have fully manifested in those patients. Possibly, neurocognitive deficits may emerge only after a certain duration of MDD, and signal worsening.Reference Allott, Fisher, Amminger, Goodall and Hetrick26 Also, as already mentioned above, the patients with MDD in some previous studies also had psychotic symptomsReference Sheffield, Karcher and Barch2 or were patients at CHR who also showed MDD symptoms.Reference Mallawaarachchi, Amminger, Farhall, Bolt, Nelson and Yuen14 The difference in the results could be driven by clinical differences between our ROD group and the other patients with MDD that took part in the previous studies.
The current study offered the possibility of direct comparison between CHR and ROD, and indicated that a widespread cognitive impairment is present in CHR compared with healthy controls, but not in ROD. Although, according to some studies, CHR represents a heterogeneous group diagnosed with a variety of common mental conditions, including anxiety and depression, the findings of the present study support the idea that CHR is a discrete condition, characterised by a cognitive impairment that is not observed in ROD. It can be argued that our CHR group also had an underlying axis 1 diagnosis that could have been present for a longer time if compared with the diagnosis reported by the ROD group. Thus, it can be argued that the longer diagnosis duration could explain the more pronounced cognitive deficits. However, this proved not to be the case, as no significant difference emerged between CHR and ROD for the duration of the axis 1 diagnosis. Also, the present results were obtained in a large sample of patients at the early stage of the illness, and who therefore were relatively spared from the prolonged illness effects. Moreover, the level of depressive symptoms was not different between the CHR and ROD groups. To make further conclusions about CHR as a separate entity in a cognitive framework, future studies looking at the longitudinal trajectory and how it may differ from the other patient groups (such as ROD and ROP) are warranted.
The current study has several limitations. First, only the baseline data were analysed in the present study. The rational of this decision was to align current work with previous studies published in the literature (e.g. Seidman et alReference Seidman, Shapiro, Stone, Woodberry, Ronzio and Cornblatt12). We agree that monitoring the development of cognitive profiles in patients at CHR and patients with ROD over time is important. As the PRONIA study also included a follow-up neurocognitive evaluation after 9 months, this will allow for further analyses that will investigate patients’ cognitive trajectories over time. However, we believe that baseline comparison of cognitive profiles of patients at CHR and, in particular, patients with ROD, who are at the very early stage of their disease, is also a necessary step for the next analyses and provides a solid ground for future contributions. Second, healthy controls were recruited throughout the whole project, whereas the patient groups (ROD, CHR, ROP) were recruited in the first 32 months, thus not representing the entire PRONIA data-set. This choice was motivated by the fact that the replication sample was deliberately kept out of the analysis for future validation purposes.
In conclusion, our study confirms that neurocognitive deficits are associated with psychosis and CHR state in all of the cognitive domains except for verbal learning. Conversely to previous studies, neurocognitive deficits seem to be absent in patients with a depressive disorder, at least in its early onset.
In general, our results point to a specificity of neurocognitive deficits as markers of vulnerability to psychosis. However, more multicentre studies using homogeneous inclusion standards and harmonised analytical procedures are needed to elucidate specific neurocognitive patterns in psychosis, CHR state and affective disorders. Ultimately, information on such distinct cognitive patterns may refine the clinical assessment in the future, and assist the early detection of mental health illnesses.
Supplementary material
Supplementary material is available online at https://doi.org/10.1192/bjp.2023.98. The normative data for the PRONIA Cognitive Battery (PCB) used in the present work can be downloaded at the following link: https://github.com/carolinabonivento/PRONIA-Cognitive-Battery-norms
Data availability
The data that support the findings of this study are available on request from the corresponding author, P.B. The data are not publicly available for ethical reasons.
Acknowledgements
The PROMIA Consortium members are as follows: Nikolaos Koutsouleris, Dominic B. Dwyer, Lana Kambeitz-Ilankovic, Anne Ruef, Alkomiet Hasan, Claudius Hoff, Ifrah Khanyaree, Aylin Melo, Susanna Muckenhuber-Sternbauer, Yanis Köhler, Ömer Öztürk, Nora Penzel, David Popovic, Adrian Rangnick, Sebastian von Saldern, Rachele Sanfelici, Moritz Spangemacher, Ana Tupac, Maria Fernanda Urquijo, Johanna Weiske, Antonia Wosgien, Camilla Krämer, Shalaila S. Haas, Rebekka Lencer, Inga Meyhoefer, Marian Surmann, Udo Dannlowski, Olga Bienek, Georg Romer, Marlene Rosen, Theresa Lichtenstein, Stephan Ruhrmann, Joseph Kambeitz, Karsten Blume, Dominika Julkowski, Nathalie Kaden, Ruth Milz, Alexandra Nikolaides, Mauro Silke Vent, Martina Wassen, Christos Pantelis, Stefan Borgwardt, Christina Andreou, André Schmidt, Anita Riecher-Rössler, Laura Egloff, Fabienne Harrisberger, Ulrike Heitz, Claudia Lenz, Letizia Leanza, Amatya Mackintosh, Renata Smieskova, Erich Studerus, Anna Walter, Sonja Widmayer, Alexandra Korda, Rachel Upthegrove, Chris Day, Sian Lowri Griffiths, Mariam Iqbal, Mirabel Pelton, Pavan Mallikarjun, Alexandra Stainton, Ashleigh Lin, Paris Lalousis, Raimo K. R. Salokangas, Alexander Denissoff, Anu Ellilä, Tiina From, Markus Heinimaa, Tuula Ilonen, Päivi Jalo, Heikki Laurikainen, Antti Luutonen, Akseli Mäkela, Janina Paju, Henri Pesonen, Reetta-Liina Säilä, Anna Toivonen, Otto Turtonen, Frauke Schultze-Lutter, Eva Meisenzahl, Sonja Botterweck, Norman Kluthausen, Gerald Antoch, Julian Caspers, Hans-Jörg Wittsack, Ana Beatriz Solana, Manuela Abraham, Timo Schirmer, Alessandro Bertolino, Linda A. Antonucci, Giulio Pergola, Ileana Andriola, Barbara Gelao, Paolo Brambilla, Carlo Altamura, Marika Belleri, Francesca Bottinelli, Adele Ferro, Marta Re, Emiliano Monzani, Maurizio Sberna, Armando D’Agostino, Lorenzo Del Fabro, Giampaolo Perna, Maria Nobile, Alessandra Alciati, Matteo Balestrieri, Carolina Bonivento, Giuseppe Cabras, Franco Fabbro, Marco Garzitto and Sara Piccin.
Author contributions
C.B., L.K.-I., E. Maggioni and P.B. were responsible for analysis or interpretation of data. C.B., L.K.-I., P.B., N.K., J.K., R.U., R.K.R.S., E. Meisenzahl, S.J.W., S.R. and S.B. were responsible for study concept and design. C.B. and L.K.-I. drafted the manuscript. All authors critically revised the manuscript for important intellectual content. C.B., E. Maggioni and P.B. were responsible for acquisition of the data and technical/data analysis. Recruitment, follow-up of the study participants, implementation of examination protocols, information technological infrastructure and data quality control were conducted by the PRONIA Consortium.
Funding
This work was supported by the European Union-FP7 project PRONIA (‘Personalised Prognostic Tools for Early Psychosis Management’; grant number 602152). J.W. was partly supported by the NARSAD Young Investigator Grant through the Brain & Behavior Research Foundation (grant number 28474). P.B. was partly supported by the Italian Ministry of Health (Ricerca Corrente 2023). E. Maggioni was partly funded by the Italian Ministry of Health (grant number GR-2019-12370616).
Declaration of interest
N.K., J.K. and R.K.R.S. are currently honorary speakers for Otsuka/Lundbeck. R.U. received grants from the Medical Research Council and the National Institute for Health Research, and personal fees from Sunovion. R.U. and L.K.-I. are members of the British Journal of Psychiatry Editorial Board, but did not take part in the review or decision-making process for this paper. The remaining authors including members of PRONIA Consortium have nothing to disclose.
eLetters
No eLetters have been published for this article.