Premorbid adjustment problems, negative symptoms, and cognitive impairment in a large international sample at clinical high risk for psychosis: Findings from the Accelerating Medicines Partnership—Schizophrenia

Abstract

Background and Hypotheses

Functional impairment often precedes attenuated psychotic symptoms (APS) in youth at clinical high risk (CHR) for psychosis. This impairment has been linked to baseline symptoms and cognition, but the roles of developmental stage of APS onset and comorbid depression/anxiety are not well understood. Accelerating Medicines Partnership—Schizophrenia’s diverse international sample and broad symptom assessment facilitate a detailed analysis of premorbid functioning.

Study Design

Premorbid adjustment and baseline symptoms and cognition were examined in 1056 individuals at CHR (n = 482 early-symptom, n = 574 late-symptom [first CHR-level APS in childhood/early adolescence vs. late adolescence/adulthood]). In the late-symptom group, growth curve models tested adjustment trends from childhood through late adolescence. Incremental relationships between adjustment, symptoms, and cognition were tested controlling for depression/anxiety.

Study Results

Premorbid social maladjustment correlated with baseline negative symptoms (r ≥ .25, P < .001). Premorbid academic maladjustment correlated with baseline negative symptoms (r ≥ .17, P ≤ .007) and cognitive impairment (r ≥ .27, P ≤ .001). Effects did not differ between early- and late-symptom subgroups (Δr ≤ .12, z ≤ 2.00, P > .05). Social adjustment deteriorated over time in participants with more baseline negative symptoms (γ = 0.14, P < .001) or cognitive impairment (γ ≥ −0.01, P < .004), even when controlling for depression/anxiety. APS were generally unrelated to premorbid functioning. Effects were consistent in the subsample residing outside North America (n = 579).

Conclusions

Relationships between premorbid adjustment, negative symptoms, and cognition were independent of depression/anxiety and developmental stage of CHR symptom onset. Baseline negative symptoms and cognitive impairment follow insidious and worsening functional problems originating in early clinical stages, prior to onset of APS.

Introduction

Young people at clinical high risk (CHR) for psychosis are typically identified based on attenuated psychotic symptoms (APS; eg, perceptual abnormalities, suspiciousness),¹ which are associated with an increased likelihood of psychotic disorders (25% within 3 years).² APS often emerge against a backdrop of longstanding functional challenges. Compared to their typically developing peers, youth at CHR report depression, anxiety, negative symptoms, and social and academic functional problems predating the emergence of APS.^3–7 Social functioning appears particularly salient, as chronic deficits and recent declines in social adjustment predict onset of psychotic disorders in CHR samples.^8–11 Many factors may account for premorbid functional trends: not only positive and negative CHR symptoms and cognition, but also nonpsychotic symptoms (depression, anxiety), sociocultural contexts, and developmental stages of CHR onset. The ongoing Accelerating Medicines Partnership—Schizophrenia (AMP® SCZ), a collaboration across 43 sites in 13 countries on 5 continents, provides a unique opportunity to explore these factors in a diverse and rigorously characterized international sample.¹²

Poor premorbid adjustment has been repeatedly linked to negative symptoms in CHR.^3,9,13,14 Poor premorbid academic adjustment has also been linked to cognitive impairment, operationalized as deficits in verbal learning or intelligence quotient (IQ), in psychotic disorders^15–20 and in one large study in CHR.⁹ However, these findings—which were primarily established in North American samples—need to be replicated internationally to assess generalizability. Additionally, it is important to replicate them using newer assessments of negative symptoms (Negative Symptom Inventory—Psychosis Risk) and cognition (Penn Computerized Cognitive Battery), which are included in AMP SCZ. If established patterns of associations between negative symptoms, cognition, and premorbid adjustment replicate in AMP SCZ, this would provide important evidence of generalizability and of newer measures’ external validity.

In contrast to the consistent evidence for negative symptoms and cognition, findings on premorbid adjustment and APS have been mixed. One study reported associations between premorbid social adjustment, unusual thought content, and suspiciousness,⁴ while another found no such relationships.³ Additional research is needed to clarify potential relationships between APS and premorbid adjustment.

Depression and anxiety are also highly prevalent in CHR, with 38% meeting diagnostic criteria for a depressive disorder and 34% meeting criteria for an anxiety disorder.²¹ CHR participants with affective disorders demonstrate more severe premorbid adjustment problems than CHR participants without such diagnoses,²² potentially predisposing them to an “affective pathway to psychosis.”²³ Furthermore, psychometric studies find strong latent associations between depression, negative symptoms, and APS.^24–27 Depression and anxiety might confound findings for premorbid adjustment in CHR (eg, negative and overgeneral memory biases associated with depression²⁸ may bias retrospective recall of functioning). These effects must be disentangled to understand whether premorbid adjustment problems are specifically linked to psychosis risk or are indicative of more general vulnerability to mental health problems.

Finally, a previous study reported that premorbid adjustment is worse in younger CHR participants, suggesting that symptoms may interfere with functioning in age-dependent ways, shaped by the developmental stage at which APS first emerge.¹¹ If symptoms emerging in different developmental stages have different functional precursors, this could be an important feature to integrate into clinical staging models (eg, interactions between developmental stage and clinical stage).²⁹ Therefore, studies examining premorbid functioning patterns in CHR would need to take into account the possible differential role of age of onset.

The present study replicates associations between premorbid adjustment, negative symptoms, and cognition in the AMP SCZ sample, and examines associations between premorbid adjustment and APS. To test incremental effects over and above nonpsychotic symptoms, analyses controlled for depression and anxiety. To consider age of onset, findings are compared in early- vs. late-symptom subgroups (first CHR-level APS appearing in childhood/early adolescence vs. late adolescence/adulthood). To differentiate functional declines from chronically poor adjustment, growth curve models estimate rates of change in adjustment from childhood to late adolescence in the late-symptom group.

Methods

Participants

Participants were 1056 community adolescents and young adults at CHR for psychosis who participated in AMP SCZ. The AMP SCZ project includes a large and diverse international sample of individuals at CHR recruited from 43 sites and 13 countries across 5 continents including North America, Australia, Europe, Asia, and South America.¹² AMP® SCZ aims to rigorously characterize a variety of clinical features and mechanisms and follow participants longitudinally to predict remission, persistence, and transition to psychosis. Complete details on the AMP® SCZ design and procedures have been reported elsewhere.¹² The study’s inclusion/exclusion criteria included age 12-30; meeting CHR criteria assessed by the PSYCHS; total lifetime antipsychotic dose <50 mg haloperidol equivalent; and no history of intellectual disability, central nervous system disorder, psychotic disorders assessed by the Structured Clinical Interview for DSM-5, or significant traumatic brain injury. All participants (or their parents/guardians for participants under 18) provided written informed consent. The current study examines baseline and Month 1 data from the AMP SCZ 3.0 data release (June 2025),³⁰ and includes all individuals in this data release with PAS data.

Procedures

Participants were assessed via clinician-administered interviews, self-report questionnaires, and tests of cognition (verbal learning, estimated IQ). Premorbid adjustment was assessed at the 1 month follow-up visit to minimize participant burden given the study’s extensive baseline assessment procedures.¹² All other data were collected at baseline.

Measures

Premorbid Adjustment

The Premorbid Adjustment Scale (PAS)^31,32 is a clinician-administered assessment that has been widely used in previous studies of premorbid adjustment in CHR.^9,11 Sociability, peer relationships, academic performance, adaptation to school, and sociosexual adaptation are scored in the developmental stages of childhood (before age 12), early adolescence (age 12-15), late adolescence (age 16-18), and adulthood (age 19+). Scores are assigned for all developmental periods up to six months before symptom onset (ie, 6 months before the first appearance of CHR-level APS). PAS scores range from 0 to 6, with higher scores indicating more problems in premorbid adjustment (ie, worse premorbid adjustment).

Participants were split into an early-symptom group and a late-symptom group based on PAS developmental periods. The early-symptom group first experienced CHR-level APS in the childhood or early adolescent periods (n = 482; 45.7%). The late-onset group first experienced CHR-level APS in the late adolescent or adulthood periods (n = 574; 54.3%). As in prior studies of the PAS in CHR, the current study examined the childhood through late adolescent periods and omitted the sociosexual adaptation subscale.^4,9,11 Factor scores for social and academic adjustment factors were calculated as means of the corresponding PAS scales (social = sociability and peer relationships; academic = academic performance, adaptation to school).¹⁶

Symptoms

Negative symptoms were assessed by the Negative Symptom Inventory-Psychosis Risk (NSI-PR),³³ a clinician-administered assessment of five consensus negative symptom domains in CHR (anhedonia, asociality, amotivation, blunted affect, and alogia). The NSI-PR evaluates text and online as well as in-person experiences, assesses both behavior and internal experience, and isolates primary and persistent negative symptoms from secondary symptoms resulting from other sources (eg, medication effects, mood episodes).³⁴ NSI-PR scores were analyzed as full-scale means and subfactor means. Following NSI-PR scoring instructions, subfactors were calculated as the mean of anhedonia, asociality, and avolition (motivation/pleasure deficits; MAP) and the mean of blunted affect and alogia (reduced expression of emotion; EXP).

APS were assessed by the Positive SYmptoms and Diagnostic Criteria for the Comprehensive Assessment of At-Risk Mental States (CAARMS) Harmonized with the SIPS (PSYCHS), a clinician-administered assessment of APS severity.¹ The PSYCHS is a standardized measure integrating symptom descriptions and diagnostic criteria from the Structured Interview for Psychosis-Risk Syndromes (SIPS) and the CAARMS to ensure compatibility across psychosis-risk assessment tools. This study examines the overall severity of APS as well as the specific symptoms of suspiciousness and unusual thought content, which have been linked to premorbid adjustment in one prior study.⁴

Depression was assessed by the Calgary Depression Scale for Schizophrenia (CDSS),³⁵ a clinician-rated assessment of depression severity. Anxiety was assessed by the Overall Anxiety Severity and Impairment Scale (OASIS), a self-report measure of anxiety severity that has been validated in clinical populations.³⁶

Cognition

Verbal learning was assessed by the Penn Language Learning Task from the Penn Computerized Neurocognitive Battery (PennCNB),³⁷ a comprehensive cognitive assessment battery administered via computer. Previous studies have found that verbal learning is a particularly relevant cognitive domain for functioning in the premorbid period.^9,15,18 In this task, participants learn a list of words, then are shown a mixture of words they memorized along with distractor words, and indicate whether each word was part of the list they initially memorized.

IQ was estimated via the 2-subtest full-scale IQ (vocabulary and matrix reasoning subtests) of the Wechsler Abbreviated Scale of Intelligence-Second Edition (WASI-II). When WASI-II translations were not available for the language spoken at a site, the corresponding subtests were substituted from the Wechsler Adult Intelligence Scale-Fourth Edition (WAIS-IV) were substituted for ages 16+ (used in 25.0% of the sample), and the Wechsler Intelligence Scale for Children-V (WISC-V) for ages up to 15 (used in 1.1% of the sample). Note that the test publisher allows these WASI and WAIS/WISC subtests to be substituted for one another.^38,39 WAIS/WISC IQ was estimated analogously to the WAIS-II 2-subtest estimate, by averaging the scaled scores on the vocabulary and matrix reasoning subtests, then converting this average score to an IQ estimate (see AMP SCZ standard operating procedures⁴⁰).

Data Analysis

Data were analyzed in the R statistical programming language, version 4.4.2.⁴¹ Descriptive analyses were conducted. Comparisons of the early- vs. late-symptom groups’ demographics and clinical variables were tested using chi-squared tests for categorical data (with Bonferroni-corrected post hoc tests to follow up significant chi-squared tests), and t-tests with effect sizes as Cohen’s d for continuous data.

To test relationships in the early- vs. late-symptom groups, Pearson correlations examined zero-order relationships between clinical variables and overall premorbid adjustment (averaged across all available developmental stages) for each adjustment factor. Partial correlations then examined the same relationships, controlling for the effects of depression and anxiety. Correlation P-values were FDR-corrected for multiple comparisons. Finally, Steiger’s z-test of independent correlations was used to test whether correlations or partial correlations differed in the early- vs. late-symptom groups.

To examine longitudinal trends in the late-symptom group’s premorbid adjustment, a series of multilevel growth curve models were estimated in {lme4} version 1.1-35.3.⁴² These analyses were limited to the late-symptom group because the early-symptom group only had data for one or two developmental periods, which would prevent models from estimating an accurate slope. Data had a multilevel structure with developmental stages nested within persons. Developmental stage was entered as a within-person variable, centered on early adolescence as the developmental midpoint of the PAS data. All other variables were entered as between-person variables. Social adjustment and academic adjustment were the dependent variables, tested in separate models. All models included a random slope for developmental period. All models included age and sex at birth as covariates. In models testing interactions (2 and 3 below), main effects represent effects at the developmental midpoint (early adolescence). To confirm main effects’ interpretability, models 2 and 3 were also calculated with main effects only (no interactions). These main-effect-only models are reported alongside models 2 and 3 in Supplementary Material A, and any differences from the interaction models are noted in the main text.

The 3 types of models were:

1. A base model estimating PAS scores as a function of developmental stage, controlling for age and sex at birth. This model establishes overall trends of premorbid adjustment in the sample. These models appear in Supplementary Material A.

2. Models testing the effects of clinical variables (negative symptoms, cognition, unusual thought content, suspiciousness, depression, and anxiety) on PAS scores, with the clinical variable and the clinical variable × developmental stage interaction entered in addition to the base model in 1 above. These models test whether the clinical variable is associated with overall poor premorbid adjustment (main effect) or trends of change in adjustment over development (interaction). These models appear in Supplementary Material A.

3. Combined models testing the incremental effect of clinical variables (negative symptoms, cognition, and positive symptoms) over and above the effect of depression and anxiety, with the main effects and interactions for the clinical variable, depression, and anxiety entered concurrently in a single model. These were the study’s primary models, and are summarized in the main text with full details reported in Supplementary Material A. Primary effects of interest were main effects and interactions for clinical variables (shown in Table 3). Although the correct family of tests for a multiple comparisons correction is debatable in this case,⁴³ we elected for a conservative approach of applying FDR-correction to the entire family of coefficients of interest (all effects reported in Table 3).

Table 3.

Growth Curve Models Estimating Premorbid Functioning, Controlling for Depression and Anxiety: Fixed Effect Coefficients (γ) and 95% CIs Shown (Full Models Shown in Supplementary A Material 3.4)

	Social adjustment		Academic adjustment
Predictor	Main effect	Interaction with developmental stage	Main effect	Interaction with developmental stage
Negative symptoms	0.31*** [0.20, 0.43]	0.15*** [0.07, 0.23]	0.21** [0.08, .033]	0.06 [−0.01, 0.14]
Motivation and pleasure	0.30*** [0.20, 0.40]	0.43** [0.06, 0.21]	0.32*** [0.21, 0.43]	0.06 [−0.01, 0.13]
Diminished expression	0.19** [−0.10, 0.28]	0.09** [0.02, 0.15]	0.07 [−0.03, 0.17]	0.03 [−0.03, 0.09]
Verbal learning	0.00 [−0.01, 0.02]	−0.02** [−0.03, −0.01]	−0.04*** [−0.06, −0.03]	−0.00 [−0.01, 0.01]
IQ	−0.00 [−0.01, 0.00]	−0.01** [−0.01, −0.00]	−0.03*** [−0.03, −0.02]	−0.00 [−0.01, 0.00]
Attenuated psychotic symptoms	−0.00 [−0.01, 0.01]	0.00 [−0.00, 0.01]	0.01 [−0.00, 0.02]	0.00 [−0.01, 0.01]
Unusual thought content	−0.06 [−0.14, 0.03]	0.02 [−0.04, 0.08]	0.03 [−0.06, 0.11]	−0.01 [−0.06, 0.05]
Suspiciousness	0.02 [−0.04, 0.08]	0.01 [−0.03, 0.06]	0.05 [−0.12, 0.01]	−0.00 [−0.04, 0.04]

Each predictor was tested in a separate model including age, sex at birth, predictor, stage, and predictor × stage interaction. Adjusted models also included main effects and interactions with stage for depression and anxiety. All models included random slopes for time. ^*P < .05; ^**P < .01; ^***P < .001. Bold indicates P < .05. All significant correlations remained significant with P ≤ .013 after FDR-correction.

Results

Descriptive Analyses

Demographic and clinical characteristics are shown in Table 1. The early-symptom group was younger than the late-symptom group, and accordingly more likely to be enrolled in school, less likely to be working, and likely to have less educational attainment. An overall difference in parental education was detected between the groups (P = .012), but Bonferroni-corrected post hoc tests found no differences on specific levels of educational attainment. The early-symptom group contained a higher proportion of females (70.3% vs. 57.4%, P < .001). Differences were also observed in regions of residence, with Bonferroni-corrected post hoc showing that the early-symptom group was more likely to reside in Australia/New Zealand (52.7% of early-symptom vs. 12.2% of late-symptom, post-hoc P < .001) and less likely to reside in Europe (6.6% vs. 23.2%, post-hoc P < .001), North America (36.5% vs. 52.4%, P < .001), and Asia (3.3% vs. 10.8%, P < .001). An overall difference was observed in the distribution of racial/ethnic identity, with the early-symptom group more likely to endorse more than one racial identity (11.6% vs. 6.4%, P = .043), more likely to endorse Hispanic or Latino identity (14.9% vs. 10.0%, P = .022), and less likely to endorse Asian identity (15.1% vs. 23.2%, P = .015), These differences in racial and ethnic identity are somewhat conflated with region of residence, complicating their interpretation in a cross-national sample. A group difference was observed in depression, which was more prevalent in the late-symptom group with a small effect size, d = 0.21, P = .001.

Table 1.

Demographic and Clinical Variables

Variable	Early-symptom N (%) or mean (SD)	Late-symptom N (%) or mean (SD)	t or χ2	d	P
N	482	574
Agea	20.11 (4.19)	22.44 (3.42)	9.72	0.64	<.001
Sex at birth (female)	338 (70.3%)	328 (57.4%)	17.94		<.001
Region of residence
Asia	16 (3.3%)	62 (10.8%)
Central/South America	3 (0.6%)	4 (0.7%)
Europe	32 (6.6%)	133 (23.2%)
North America	176 (36.5%)	301 (52.4%)
Australia/New Zealand	254 (52.7%)	70 (12.2%)	221.81		<.001
Immigration status
Nonimmigrant	331 (69.0%)	396 (69.8%)
1st gen. immigrant (moved age < 5)	51 (10.6%)	58 (10.2%)
1st gen. immigrant (moved age > 5)	22 (4.6%)	23 (4.1%)
2nd gen. immigrant	76 (15.8%)	90 (15.9%)	0.23		.972
Racial identity
Asian	73 (15.1%)	133 (23.2%)
Black or African American	27 (5.6%)	57 (9.9%)
White	303 (62.9%)	327 (57.0%)
Multiple	56 (11.6%)	37 (6.4%)
Other	10 (2.1%)	8 (1.4%)	27.36		<.001
Hispanic/Latino	48 (10.0%)	85 (14.9%)	5.305		.021
Employment status
Not working	255 (53.0%)	255 (44.7%)
Part-time	162 (33.7%)	216 (37.9%)
Full-time	64 (13.3%)	99 (17.4%)	7.74		.021
Current student	349 (72.6%)	352 (61.6%)	13.49		<.001
Participant education
Less than seven years of school	19 (4.0%)	2 (0.4%)
Junior high school	44 (9.1%)	10 (1.8%)
Some high school/secondary school	118 (24.5)	77 (13.5%)
Completed high school	115 (23.9%)	151 (26.4)
Partial college training	88 (18.3%)	173 (30.3%)
College or university education	85 (17.7%)	136 (23.8%)
Graduate or professional training	12 (2.5%)	22(3.9%)	83.96		<.001
Parental education
Less than seven years of school	1 (0.2%)	8 (1.4%)
Junior high school	6 (1.3%)	11 (2.0%)
Some high school/secondary school	41 (8.6%)	48 (8.6%)
Completed high school	96 (20.2%)	78 (13.9%)
Partial college training	66 (13.9%)	65 (11.6%)
College or university education	170 (35.8%)	205 (36.5%)
Graduate or professional training	95 (20.0%)	146 (26.0%)	16.36		.012
Negative symptoms (NSI-PR overall score)b	1.12 (0.76)	1.15 (0.78)	0.7	0.03	.483
Motivation and pleasure	1.86 (0.85)	1.84 (0.88)	0.25	0.02	.804
Diminished expression	0.71 (0.89)	0.76 (0.95)	0.92	0.05	.357
Attenuated psychotic symptoms (PSYCHS overall score)	20.60 (8.39)	21.23 (9.26)	1.16	0.07	.245
Unusual thought content	3.55 (0.99)	3.59 (1.03)	0.68	0.03	.496
Suspiciousness	2.87 (1.37)	2.88 (1.49)	0.17	0.01	.869
Verbal learning (PennCNB)	28.56 (6.73)	28.69 (6.57)	0.3	0.01	.767
IQ (WASI-II /WAIS-IV/WISC-V)	110.63 (13.2)	109.80 (14.16)	0.95	0.06	.342
Depression (CDSS)	5.68 (4.19)	6.64 (4.62)	3.47	0.21	.001
Anxiety (OASIS)	8.55 (3.88)	8.89 (4.03)	1.38	0.08	.169

^aNote that groups were defined by the developmental stages at which attenuated psychotic symptoms first appeared, not by age at baseline.

^bPer the instruments’ scoring instructions, the NSI-PR overall score is calculated as a mean while total scores for PSYCHS, CDSS, and OASIS are calculated as sums.

Correlational Results in Early- and Late-Symptom Groups

As shown in Table 2, in both early- and late-symptom groups, premorbid social adjustment problems correlated with baseline negative symptoms (r ≥ .21, P < .001) and depression (r ≥ .11, P ≤ .011). In both early- and late-symptom groups, premorbid academic adjustment problems correlated with baseline negative symptoms (r ≥ .19, P < .001), poorer baseline verbal learning (r ≥ −0.26, P < .001), and lower baseline IQ (r ≥ −0.28, P < .001). These correlations remained significant when controlling for the effects of depression and anxiety.

Table 2.

Correlations Between Overall Premorbid Adjustment and Clinical Variables in Individuals with Early- vs. Late-Onset of Attenuated Psychotic Symptoms

	Social adjustment						Academic adjustment
	Zero-order correlations			Partial correlations controlling for depression/anxiety			Zero-order correlations			Partial correlations controlling for depression/anxiety
Variable	Early-onset	Late-onset	Diff.	Early-onset	Late-onset	Diff.	Early-onset	Late-onset	Diff.	Early-onset	Late-onset	Diff.
Negative symptoms	0.25***	0.25***	0.00	0.21***	0.23***	−0.01	0.20***	0.19***	0.01	0.19***	0.16***	0.03
Motivation and pleasure	0.25***	0.26***	0.01	0.21***	0.23***	−0.02	0.23***	0.28***	−0.04	0.22***	0.24***	−0.01
Diminished expression	0.18***	0.19***	0.01	0.16***	0.18***	−0.02	0.15**	0.10*	0.05	0.13**	0.08	0.05
Attenuated psychotic symptoms	0.09	0.02	0.06	0.04	−0.01	0.05	0.07	0.09*	0.01	0.05	0.05	0.00
Unusual thought content	0.07	−0.02	0.09	0.03	−0.05	0.09	0.03	0.05	−0.02	0.00	0.01	−0.00
Suspiciousness	0.08	0.07	0.00	0.03	0.04	−0.00	0.04	0.09*	−0.05	0.01	0.06	−0.04
Verbal learning (PennCNB)	−0.07	−0.02	0.05	−0.09	−0.01	−0.08	−0.26***	−0.28***	0.01	−0.27***	−0.26***	−0.01
IQ (WASI-II or WAIS-IV)	−0.03	−0.06	0.03	−0.03	−0.04	0.01	−0.28***	−0.37***	0.09	−0.29***	−0.36***	0.07
Depression (CDSS)	0.15**	0.11*	0.03	-	-	-	0.06	0.18***	−0.12*	-	-	-
Anxiety (OASIS)	0.08	0.10*	0.02	-	-	-	0.06	0.06	0.00	-	-	-

Abbreviations: Early-onset = onset of APS in childhood or early adolescence; late-onset = onset of APS in late adolescence or adulthood; Diff. = difference. Developmental periods were defined by the Premorbid Adjustment Scale. Differences between early- and late-symptom correlations, shown in italics, were tested via Steiger’s z-test of differences in independent correlations. ^*P < .05; ^**P < .01; ^***P < .001. Bold indicates P < .05. Correlation P-values were FDR-corrected for multiple comparisons.

As shown in Table 2, only one difference in correlations was observed between the early- and late-symptom groups. Depression was more strongly correlated with academic functioning in the late-onset than the early-onset group, early P = .06, late P = .18, Δ = 0.12, Z = 1.65, P = .045.

Developmental Trends in Late-Symptom Group

Base Model

In a base model examining the relationship between PAS scores and developmental stage (Model 1 in Data Analysis above, reported in Supplementary Material A 3.2), academic functioning deteriorated over the course of development, γ = 0.25 [0.20, 0.31], P < .001, whereas social functioning did not, γ = 0.01 [−0.05, 0.07], P = .654.

Depression and Anxiety

To confirm the relevance of depression and anxiety for premorbid adjustment trends, growth curve models tested relationships between depression, anxiety, and premorbid adjustment (Model 2 in Data Analysis above, reported in Supplementary Material A 3.3.4). Participants with more severe depression at baseline reported poorer overall premorbid social adjustment, γ = 0.03 [0.01, 0.05], P = .002, and academic adjustment, γ = 0.05 [0.03, 0.07], P < .001. Participants with more severe anxiety at baseline reported poorer overall premorbid social adjustment, anxiety γ = 0.03 [0.01, 0.05], P = .005, but not academic adjustment (P = .059). Interactions between depression/anxiety and developmental stage were nonsignificant (p≥.177).

Primary models reported below included depression and anxiety as covariates (Model 3 in Data Analysis above). To allow results to be directly compared to previous growth curve models that did not control for depression/anxiety,⁹ models were also calculated without controlling for depression and anxiety (Model 2 in Data Analysis above, reported in Supplementary Material A 3.3). Additionally, models were calculated with main effects only (reported alongside primary models in Supplementary Material A 3.4), to confirm that main effects remained significant when interactions were omitted from the models. In both cases, results were very similar to the primary models, and any differences are noted in the text.

Negative Symptoms

As summarized in Table 3 and Figure 1, and shown in full in Supplementary Material A 3.4.1, participants with more severe negative symptoms at baseline reported poorer overall social adjustment, γ = 0.32 [0.21, 0.44], P < .001, poorer overall academic adjustment, γ = 0.21 [0.08, 0.33], P = .001, and trends of worsening social adjustment over time, γ = 0.15 [0.07, 0.23], P < .001. Effects were notably stronger for the Motivation and Pleasure (MAP) symptom factor than for the Diminished Expression (EXP) symptom factor (see Table 3 and Figure 1).

Figure 1.

Premorbid adjustment as a function of negative symptoms. Panels show retrospective change over time in premorbid social or academic adjustment as a function of overall negative symptoms (A, B), motivation and pleasure deficits (C, D), and emotion expression deficits (E, F). Colored lines indicate the predicted growth curve at 1 SD, mean, and −1 SD of the relevant symptom. All models controlled for age, sex at birth, depression, and anxiety. Higher scores (plotted down) indicate more severe premorbid adjustment problems assessed by the premorbid adjustment scale. Significance levels of main effects and interactions are noted at ^*P < .05, ^**P < .01, ^***P < .001. n = 574 individuals at clinical high risk for psychosis with late symptom onset (first attenuated psychotic symptoms appearing in late adolescence or adulthood).

Cognition

As summarized in Table 3 and Figure 2, and shown in full in Supplementary Material A 3.4.2, participants with worse verbal learning reported poorer overall academic adjustment, γ = −0.04 [−0.06, −0.03], P < .001, and trends of deteriorating social adjustment over time, γ = −0.02 [−0.03, −0.01], P = .004. Similarly, participants with lower IQ reported poorer overall academic adjustment, γ = −0.03 [−0.03, −0.02], P < .001, and trends of deteriorating social adjustment over time, γ = −0.01 [−0.01, −0.00], P = .003.

Figure 2.

Premorbid adjustment as a function of cognition. Panels show retrospective change over time in premorbid social or academic adjustment as a function of verbal learning (A, B) and IQ (C, D). Colored lines indicate the predicted growth curve at 1 SD, mean, and −1 SD of the relevant cognitive variable. All models controlled for age, sex at birth, depression, and anxiety. Higher scores (plotted down) indicate more severe premorbid adjustment problems assessed by the premorbid adjustment scale. Significance levels of main effects and interactions are noted at ^*P < .05, ^**P < .01, ^***P < .001. n = 574 individuals at clinical high risk for psychosis with late symptom onset (first attenuated psychotic symptoms appearing in late adolescence or adulthood).

Attenuated Psychotic Symptoms

As summarized in Table 3 and shown in full in Supplementary Material A 3.4.3, neither overall baseline severity of APS nor unusual thought content was associated with premorbid social or academic adjustment (all main effect and interaction P ≥ .127). In models uncontrolled for depression and anxiety (Supplementary Material A 3.3.3), overall APS severity, γ = 0.01 [0.00, 0.02], P = .007, and suspiciousness, γ = 0.07 [0.01, 0.13], P = .028, were both associated with poorer overall premorbid academic adjustment. However, these effects were nonsignificant when controlling for depression and anxiety.

Subgroup Analysis in Non-North American Participants

All analyses above were repeated in the subgroup of participants residing outside North America (n = 579), shown in Supplementary Material B. Nearly all effects were significant in the same direction in this subgroup. Only two substantive differences were observed. First, Z-tests of differences in partial correlations (controlling for depression and anxiety) between the early-symptom and late-symptom groups (Supplementary Material B 2.3) were significant for associations between negative symptoms and social adjustment problems (early part. r = 0.14, late part. r = 0.31, Δr = 0.17, z = 2.08, P = .037) and between unusual thought content and social adjustment problems (early part. r = 0.08, late part. r = −0.14, Δr = 0.22, z = 2.60, P = .009). Second, counterintuitively, growth curve models found that unusual thought content was associated with less premorbid social adjustment problems, γ = −0.14 [−0.27, −0.01], P = .031 (Supplementary Material B 3.4.3), although this was only observed in the model controlling for depression and anxiety, and not in models without these covariates (P = .124; Supplementary Material B 3.3.3).

Discussion

In a large, diverse international sample of individuals at CHR for psychosis, relationships between negative symptoms, cognition, and premorbid adjustment proved to be robust. A consistent pattern emerged: poor premorbid social adjustment preceded baseline negative symptoms, while poor premorbid academic adjustment preceded baseline negative symptoms and cognitive impairment. These effects were consistent across multiple analytical approaches and independent of age of symptom onset and comorbid depression and anxiety. In the late-symptom group, growth curve models found that longitudinal declines in social adjustment preceded more severe baseline negative symptoms and cognitive impairment. Participants experiencing these declines may be of particular clinical concern, as decline in social adjustment often precedes the onset of psychotic disorders.^8–11

Replicating and Extending Prior Work with New Measures in an International Sample

Associations between clinical variables and premorbid adjustment were consistent with previous studies assessing negative symptoms using the SIPS^3,9,13,14 and verbal learning using the MATRICS Consensus Cognitive Battery (MCCB).⁹ Growth curve modeling results for negative symptoms and verbal learning closely matched those from a similar growth curve modeling study of the NAPLS-3 cohort using the SIPS and MCCB.⁹ Of six significant effects for negative symptoms and verbal learning observed in that study, five replicated in the current study. This provides evidence for: (1) the robustness of these effects across multiple sociocultural contexts, assessment measures, and developmental stages; (2) the validity of the NSI-PR and Penn CNB as measures linked to developmental functioning patterns; and (3) the utility of growth curve modeling as a method to analyze retrospective functional data.

Findings were largely consistent in participants residing outside North America (n = 579), suggesting that links between premorbid adjustment, negative symptoms, and cognitive impairment are generalizable internationally. Minor differences in the association between premorbid social functioning and clinical presentation were found in participants residing outside North America. Regional differences in norms and attitudes regarding CHR have previously been reported. For example, some Tunisian primary care physicians attributed the CHR state to loneliness (22.4%), character weaknesses (21.4%), God’s will (15.9%), or witchcraft (6.6%).⁴⁴ The present study suggests that there may also be subtle cultural variation in the developmental dynamics preceding CHR, perhaps suggesting that emerging symptoms impact functioning differently in the context of different cultural norms regarding development of autonomy, identity, peer, and family relationships in adolescence.^45,46 Future cross-cultural research could explore these factors in greater detail.

SIPS negative symptoms were previously reported to be associated with deteriorating academic adjustment over time,⁹ and this was the only effect not to replicate in the current study. SIPS negative symptoms contain a meaningful amount of variance from a general symptom factor associated with functional impairment.^25,47 The NSI-PR was designed to more specifically isolate negative symptoms,³³ and may contain less general factor variance. Additional factor analytic research on the NSI-PR could support or disprove this speculation.

Negative Symptoms

Negative symptoms were most strongly associated with poor premorbid adjustment, both in terms of overall adjustment problems and in worsening adjustment from childhood through late adolescence. Effects seemed to be driven by deficits in motivation and pleasure (anhedonia, asociality, and avolition) more than by emotional expression. Premorbid functional problems and baseline motivation and pleasure deficits may reflect worsening manifestations of a shared underlying dimension. Issues with anhedonia, amotivation, and asociality may progress from subtler problems that begin to impact functioning in earlier clinical stages to more significant problems—diagnosable as negative symptoms and significantly impacting functioning—by the time a CHR syndrome is identified. The HiTOP detachment spectrum (a transdiagnostic psychopathology dimension capturing social withdrawal, emotional constriction, and interpersonal disinterest) may be a candidate to account for underlying shared mechanisms between negative symptoms and premorbid adjustment problems.⁴⁸ Emerging problems in this spectrum may have early and ongoing effects on functioning in the premorbid period.

Depression and Anxiety

Depression and anxiety were prevalent in the current sample, with mean depression scores around the clinical cutoff of 6 (early-symptom group: 5.84; late-symptom group: 6.88)³⁵ and mean anxiety scores above the clinical cutoff of 8 (early-symptom group: 8.55; late-symptom group: 8.89).³⁶ Current depression and anxiety also had meaningful associations with premorbid functioning. Factor-analytic research has highlighted psychometric relationships between depression, anxiety, APS, and functional impairment in CHR.^24,26,27 This has been found even when depression has been assessed by the CDSS²⁵ (used in the current study), which was designed to assess depression symptoms that do not overlap with negative symptoms in schizophrenia.³⁵ In the present study, effects for negative symptoms and cognitive impairment were independent of effects for depression and anxiety. Developmental continuities between premorbid functioning and current negative symptoms and cognition were not merely artifacts of depression and anxiety. They reflected unique and specific relationships between premorbid functioning trends and characteristics of the CHR state.

Attenuated Psychotic Symptoms

APS were largely unrelated to premorbid functioning. While APS are the criteria for determination of CHR status and the most studied symptoms in CHR, negative symptoms and cognitive problems were more related to longstanding patterns of functional impairment in this sample. Follow-back studies of people with psychotic disorders find that depression, anxiety, and functional impairment are as common as APS prior to psychosis⁴⁹ and are often the earliest observable symptoms in people who later develop psychosis.⁷ Together with the present findings, this would be consistent with a pattern of depression, anxiety, and negative symptoms appearing and causing functional impairment earlier in development, prior to emergence of APS. This pattern lends further support to transdiagnostic clinical staging models^6,29 and historical theories that symptoms develop hierarchically from nonspecific dysthymic states to specific psychotic symptoms.⁵⁰

Retrospective Functional Assessment

A variety of clinical research literatures use retrospective or follow-back assessment,^7,49,51–53 including the original follow-back research on the psychosis prodrome that gave rise to the CHR paradigm.^54,55 One potential concern with this approach is that retrospective reporting may be colored by recall biases associated with current symptom severity.²⁸ Previous research has demonstrated that PAS scores correspond to actual functioning reported during the relevant developmental periods,^56,57 that current symptom severity does not affect accuracy of recall for premorbid adjustment in schizophrenia,⁵⁷ and that baseline social and academic PAS scores predict 10-year functional outcomes when controlling for baseline variables.⁵⁸ The current study adds to this body of evidence for the validity of PAS scores independent of current symptom severity, by finding that PAS scores assessed one month apart from symptoms still show the expected pattern of associations with symptoms.

Limitations and Future Directions

This study was limited by its use of linear growth curve models, which can only estimate linear trends across developmental periods. Model complexity was constrained by the limited number of within-person data points provided by the PAS. Despite these limitations, multilevel growth curve modeling is still valuable for its ability to estimate trends while accounting for clustered residuals in longitudinal data. Similarly, although the diverse international sample is a notable strength of the AMP® SCZ study, we observed regional and racial differences in age of symptom onset according to the PAS. These may reflect regional differences in study recruitment or in pathways to mental healthcare, for example, experiences may be classified as mental health symptoms earlier or later in different sociocultural contexts and healthcare systems. It may be useful to modify retrospective assessments of functioning for international applications and cultural differences in expected functional trends during development. For instance, the PAS assumes that children and adolescents are mandated to attend school full-time, which is a culturally-bounded assumption. A more comprehensive IQ assessment available in all languages (eg, WAIS/WISC GAI) would also have provided stronger evidence of IQ measurement equivalence across countries. Additionally, anxiety was assessed via self-report while other measures were assessed by clinician-administered interviews, and mode of administration may have altered the strength of observed effects. Parent- or teacher-report of premorbid adjustment, and of depression/anxiety/negative symptom severity in premorbid developmental stages, would also be valuable for eliminating possible roles of recall biases. Additionally, the current study’s models were not well suited to test interactive effects between social x academic functioning, which would be an important future extension of this work. Finally, long-term outcome data are not yet available from AMP® SCZ, constraining potential inferences about premorbid adjustment associated with psychotic disorder incidence or progression/remission/persistence of the CHR state.

Conclusion

This study highlights concerning patterns of functional deterioration preceding identification of a CHR syndrome. Negative symptoms and cognitive problems should be understood in the context of long-term functional challenges in young people’s lives.

Funding

The Accelerating Medicines Partnership Schizophrenia (AMP SCZ) is a public-private partnership managed by the Foundation for the National Institutes of Health. The AMP SCZ research program is supported by contributions from the AMP SCZ public and private partners, including NIMH (U24MH124629, U01MH124631, and U01MH124639) and the Wellcome Trust (220664/Z/20/Z and 220664/A/20/Z). The support of the NHMRC CRE PREEMPT (1198304) is acknowledged. Preparation of this manuscript was supported by a NARSAD Frederick & Alice Coles and Thomas & Nancy Coles Young Investigator Award (32738 to HRC).

Conflicts of Interest

None declared.