Combinations and Temporal Associations Among Precursor Symptoms Before a First Episode of Psychosis

Abstract

Background and Hypothesis

Symptoms that precede a first episode of psychosis (FEP) can ideally be targeted by early intervention services with the aim of preventing or delaying psychosis onset. However, these precursor symptoms emerge in combinations and sequences that do not rest fully within traditional diagnostic categories. To advance our understanding of illness trajectories preceding FEP, we aimed to investigate combinations and temporal associations among precursor symptoms.

Study Design

Participants were from PEPP-Montréal, a catchment-based early intervention program for FEP. Through semistructured interviews, collateral from relatives, and a review of health and social records, we retrospectively measured the presence or absence of 29 precursor symptoms, including 9 subthreshold psychotic and 20 nonpsychotic symptoms. Sequences of symptoms were derived from the timing of the first precursor symptom relative to the onset of FEP.

Study Results

The sample included 390 participants (68% men; age range: 14–35 years). Combinations of precursor symptoms most frequently featured depression, anxiety, and substance use. Of 256 possible pairs of initial and subsequent precursor symptoms, many had asymmetrical associations: eg, when the first symptom was suspiciousness, the incidence rate ratio (IRR) of subsequent anxiety was 3.40 (95% confidence interval [CI]: 1.79, 6.46), but when the first symptom was anxiety, the IRR of subsequent suspiciousness was 1.15 (95% CI: 0.77, 1.73).

Conclusions

A detailed examination of precursor symptoms reveals diverse clinical profiles that cut across diagnostic categories and evolve longitudinally prior to FEP. Their identification may contribute to risk assessments and provide insights into the mechanisms of illness progression.

Introduction

The first episode of psychosis (FEP) is generally preceded by precursor signs and symptoms, emerging over a period of days to years before the onset of threshold-level psychosis.^1–4 Precursor symptoms—both those that resolve before the onset of psychosis (outpost symptoms) and those that are contiguous with it (representing the prodrome)—can present as part of specific syndromes, such as subthreshold psychosis or depression.³ The idea that individuals experiencing a precursor stage could be identified and followed prospectively to prevent or delay onset of a psychotic disorder has motivated the development of early intervention services for individuals at clinical high risk for psychosis,^5,6 initially pioneered in Australia and now spreading globally.^7–9

A series of landmark retrospective cohort studies, most prominently Hafner and colleagues’ ABC project,¹⁰ have identified the prevalence and timing of precursor symptoms—noting, for example, the co-occurrence of subthreshold psychotic and nonpsychotic symptoms as part of pre-onset trajectories to an FEP.^3,4,11–13 These retrospective studies also showed that suspiciousness is particularly common before FEP,² and that it is associated with faster progression toward FEP.¹⁴ Precursor symptoms before an FEP also combine together in complex ways that do not rest fully within traditional diagnostic categories. Indeed, individual precursor symptoms can be conceived as powerful psychological experiences that give rise to other symptoms and that participate in the progression of illness, as suggested by lived experience testimonies¹⁵ and network theories of psychopathology.¹⁶

Building on these findings, our understanding of the precursor stage may be advanced by systematically investigating the combinations of precursor symptoms across syndromes. In the context of early intervention services, this symptom-based approach could strengthen our capacity to personalize risk assessments and to target specific symptoms that potentially underlie illness progression.

Having established the prevalence of pre-onset subthreshold psychotic syndromes in our FEP cohort,¹⁷ we have since reported on the distribution and relationship between nonpsychotic and subthreshold psychotic symptoms,¹¹ prospective clinical outcomes during FEP treatment,^18,19 and neurobiological markers.^20,21 More recent work documenting the rapidity of illness progression across a range of initial symptoms¹⁴ and the relationship between pre-onset self-harm and longitudinal outcomes²² has suggested the need to undertake a systematic and comprehensive examination of how precursor symptoms combine together and are sequenced over time. Accordingly, the present study aimed to use the same dataset to (1) describe common combinations of precursor symptoms prior to FEP; and (2) evaluate temporal associations between pairs of precursor symptoms, in order to determine whether the content of the first precursor symptom is associated with the content of subsequent precursor symptoms. We employ, for the first time, combination analyses and interactive data visualization of 29 precursor symptoms and 256 temporal pairs thereof.

Methods

Participants

Participants were from PEPP-Montréal, an early intervention service for individuals with FEP.²³ PEPP-Montréal is the only such service within a catchment area of ~300 000 individuals in the southwest of Montreal, Canada. Participants were recruited from consecutive admissions into PEPP-Montréal between 2003 and 2017. Eligibility criteria for study participation were the criteria for admissibility into the clinic: (1) affective or nonaffective psychotic disorder based on SCID-IV criteria,²⁴ confirmed by a senior psychiatrist (R.J., A.K.M., or J.L.S.) and not explained by substance use alone; (2) having received no more than 30 days of antipsychotic medication; (3) IQ of at least 70; (4) no organic mental disorder such as epilepsy; and (5) age between 14 and 35 years. Study participants provided written informed consent (or assent if they were underage). All procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. All procedures involving patients were approved by the Douglas Research Centre’s ethics board.

Measures

After their admission to PEPP-Montréal, participants underwent systematic assessments of their precursor symptoms using the Circumstances of Onset and Relapse Schedule and its module, the Topography of Psychotic Episode.²⁵ This instrument retrospectively evaluates 27 precursor symptoms of FEP (supplementary table S1) based on semistructured interviews with participants and family members to determine whether each symptom was ever present or not (dichotomously) during the precursor stage. Measures also include the ages at onset of the first precursor symptom and at onset of FEP. Anchor points (eg, “my 16th birthday” or “during 8th grade”) were used to estimate the timing of events. Interviews were conducted by trained research assistants and complemented by a detailed review of all available medical and social records to identify the presence of precursor symptoms and their timing. Research assistants then systematically synthesized the data from interviews and medical and social records into a structured timeline of the participant’s precursor period. This timeline was reviewed during regular meetings of the research team chaired by a senior psychiatrist (R.J., A.K.M., or J.L.S.), and discrepancies were resolved by consensus. The instrument, which is derived from the Interview for the Retrospective Assessment of the Onset of Schizophrenia,²⁶ was found to have good interrater reliability at PEPP-Montréal (intraclass correlation coefficient, 0.81–0.98).²⁵

As assessed by an international panel of experts,¹⁷ these 27 precursor symptoms were organized into 2 categories: subthreshold psychotic symptoms (9 symptoms) and nonpsychotic symptoms (18 symptoms). We previously showed that having subthreshold psychotic symptoms in the precursor stage was associated with poorer outcomes after 1 year of treatment in PEPP-Montréal, supporting the predictive validity of these measures.¹⁸ In the current study, we have augmented the 18 precursor nonpsychotic symptoms with 2 additional precursor ones, alcohol use problems (eg, alcohol-related functional impairments) and any other substance use (eg, cannabis, stimulants, opioids), given their inclusion in transdiagnostic clinical staging models of mental illness.²⁷ For each participant, all identified precursor symptoms were then labeled as either the first identifiable symptom or as any of the subsequent symptoms until onset of psychosis.

Statistical Analysis

Analyses were conducted in R version 4.1.2 (R Foundation for Statistical Computing) between June 15, 2021 and September 1, 2023. Participants who did not complete assessments of precursor symptoms were removed from analyses. Participants included and excluded from analyses for this reason were compared descriptively.

Combinations of Precursor Symptoms

First, to describe the most common combinations of precursor symptoms, we generated UpSet plots using the ComplexHeatmap package.²⁸ Like Venn diagrams, UpSet plots illustrate combinations (single, pairs, trios, etc.) within a set of features (ie, precursor symptoms). Relative to Venn diagrams, UpSet plots are better suited to representing large numbers of combinations. Nonetheless, because of the computational cost of combining 29 symptoms (and since many symptoms were infrequently reported in the sample), we limited the combination analysis to the 10 most common symptoms in the dataset. We explored group differences by generating UpSet plots separately for male and female participants and for participants who received diagnoses of affective and nonaffective psychosis.

Temporal Associations Between First and Subsequent Precursor Symptoms

Second, to unpack combinations of symptoms over time, we explored their temporal associations using Poisson regressions with sandwich estimators for standard errors.²⁹ Within participants, all identified precursor symptoms were either categorized as the first symptom or one of the subsequent symptoms. The modified Poisson regression method is adequate for examining dichotomous events,²⁹ and we chose this approach over Cox models given the absence of attrition (censoring) over time stemming from the retrospective design. For all first symptoms observed in the dataset, we estimated the incidence rate ratio (IRR) for specific subsequent symptoms. To do so, the model’s predictor was a dichotomous variable indicating, for each participant, whether X symptom was the first symptom (eg, suspiciousness as first symptom: Yes/No), and the outcome was a dichotomous variable indicating whether Y symptom was a subsequent symptom (eg, anxiety as one of the subsequent symptoms: Yes/No). The timeframe in the denominator of the IRR was the interval in years between the onset of the first symptom and the onset of FEP. In addition to the 29 precursor symptoms described above, we included 2 aggregate categories: any subthreshold psychotic symptom (Yes/No) and any nonpsychotic symptom (Yes/No). We explored interactions with sex and diagnosis of affective vs nonaffective psychosis. Significant interactions were probed using estimated marginal means.³⁰

Overall, the questions that these analyses sought to answer can be summarized by taking the pair of suspiciousness and anxiety as an example: “Is a first symptom of suspiciousness, relative to any other possible first symptom, associated with a higher or lower likelihood of subsequent anxiety?” In this example, an IRR <1 would indicate that a first symptom of suspiciousness is associated with a lower likelihood of subsequent anxiety, whereas an IRR >1 indicates a higher likelihood thereof. We examined this question across all 31 precursor symptoms, with each precursor alternately included as first or subsequent symptom, leading to 60 possible combinations per symptom, or 930 possible combinations in total. To limit the identifiability of participants, however, we only computed models when pairs of first and subsequent symptoms were observed in at least 5 persons,¹⁴ leading to a total of 256 symptom pairs investigated out of the 930 possible combinations. These temporal associations are presented in an interactive visualization inspired by Plana-Ripoll et al.³¹ Given the hypothesis-generating aim of the study, we report both uncorrected 95% confidence intervals (CIs) and P values adjusted for false discovery rate. We adjusted for false discovery rate using the Benjamini and Hochberg method³² as a function of the number of associations involving the symptom of interest: eg, of 60 possible sequences of symptoms involving depression as the first or subsequent symptom, we examined 41 sequences that had at least 5 observations, leading to adjustment for 41 comparisons. Statistical significance was defined as adjusted P values <.05.

Results

Sample Characteristics

Of 780 individuals admitted to PEPP-Montréal, 626 (80.3%) consented to join the research cohort. Of them, 390 (62.3%) completed the assessment of precursor symptoms and were thus included in the current analyses; their characteristics were similar to those of excluded participants, except for a slightly lower proportion of individuals who did not complete high school (table 1).

Table 1.

Characteristics of Included and Excluded Participants

	Included Participants (N = 390)		Excluded Participants (N = 235)
	N	%	N	%
Sex
Female	123	31.5	66	28.1
Male	266	68.2	168	71.5
Racialized group
White	239	63.4	127	60.5
Black	45	11.9	33	15.7
Asian	34	9.0	19	9.1
Arab	21	5.6	8	3.8
Other	19	5.0	10	4.8
Multiple	19	5.0	13	6.2
Educational attainment
Less than high school	178	47.2	118	55.4
High school or higher	199	52.8	95	44.6
Primary diagnosis for first episode of psychosis:
Affective psychosis	109	28.2	68	30.9
Nonaffective psychosis	277	71.8	152	69.1
Secondary diagnosis of substance use disorder
Yes	200	53.6	101	51.5
No	173	46.4	95	48.5

Combinations of Precursor Symptoms

Participants (n = 390) had a median count of 9 precursor symptoms (interquartile range: 5–12) over a median duration of the precursor stage of 3.83 years (interquartile range: 1.17–8.51). When examining mutually exclusive symptom combinations, only 6 combinations were shared by ≥5 participants, with the most common combination involving the presence of all top-10 precursors in 4% of participants (supplementary figure S1). To identify more prevalent patterns, we next considered nonexclusive symptom combinations, such that, eg, individuals with “depression plus anxiety” combination would also count among individuals with the “depression” combination. (In this approach, single-symptom combinations are particularly frequent since they are less restrictive than multi-symptom combinations.)

In the total sample (figure 1), the most common nonpsychotic symptom, and most common “combination” overall, was depression (74% of participants). The most common subthreshold psychotic symptom was suspiciousness (46%). The most common pair and trio of precursor symptoms, respectively, were depression plus anxiety (56%), and depression plus anxiety plus social withdrawal (38%). Combinations of precursor symptoms by sex and diagnosis are in supplementary figures S2–S5. Among female participants, the 3 most common combinations (including single symptoms) were depression (83%), anxiety (78%), and depression plus anxiety (68%), whereas among male participants, they were substance use (75%), depression (70%), and impaired role functioning (65%). Among participants with affective psychosis, the 3 most common precursor combinations were depression (87%), anxiety (72%), and depression plus anxiety (66%), whereas among those with nonaffective psychosis, they were depression (69%), anxiety (68%), and substance use (68%).

Fig. 1.

Prevalence of the most common precursor symptom combinations. UpSet plot indicating the relative fraction (frequency) of the 50 most common combinations of precursor symptoms. The relative fraction of each combination is illustrated by bar plots in the upper part of the figure. The list of precursor symptoms, limited to the 10 most common symptoms, is on the left. Combinations of these symptoms are illustrated by connected dots. Bar plots on the right-hand side indicate the number of participants who had each of the individual precursor symptoms. Combinations are intersecting (not mutually exclusive).

Temporal Associations Between First and Subsequent Precursor Symptoms

The 256 temporal associations between pairs of precursor symptoms are summarized in figure 2. Associations varied as a function of the composition and order of symptom pairs. In some cases, such as suspiciousness, impaired role functioning, or any subthreshold psychotic symptom as the first symptom vs other possible first symptoms, all significant associations indicated a higher likelihood of subsequent symptoms (IRR >1). In other cases, such as anxiety as the first symptom, both directions of associations were found (lower and higher likelihood of subsequent symptoms); and after a nonpsychotic symptom as the first symptom, all significant associations indicated lower likelihood of subsequent symptoms (IRR <1). Individual associations are presented at https://vincepaquin.shinyapps.io/TOPEdata and illustrated below by focusing on 2 sets of interest: pairs involving the most common nonpsychotic and subthreshold psychotic symptoms, respectively, depression and suspiciousness.

Fig. 2.

Summary of temporal associations between pairs of precursor symptoms. For each symptom listed on the left, horizontal bars indicate the number of associations with other precursor symptoms where the listed symptom was either the first symptom in the sequence (right panel) or the subsequent symptom (left panel). Bars indicate associations that were (1) negative (incidence rate ratio [IRR] <1), indicating a lower likelihood of observing this sequence of symptoms, (2) not statistically significant, or (3) positive (IRR >1), indicating a higher likelihood of observing this sequence of symptoms. Associations were not computed for sequences of symptoms observed in <5 participants. Statistical significance: P values <.05 after adjustment for false discovery rate among associations involving the symptom of focus. Note: NPS, nonpsychotic symptom; STPS, subthreshold psychotic symptom.

Pairs involving depression as the first or subsequent symptom are presented in figure 3. Symptoms paired with depression are listed on the left of the figure, and their associations with depression are indicated by points and horizontal lines. Triangles represent associations where depression is the first symptom of the pair, while circles represent associations where depression is the subsequent symptom. Associations of depression with 4 other precursors (elated mood, suspiciousness, nonpsychotic symptoms, and subthreshold psychotic symptoms) differed depending on whether depression was identified as the first vs subsequent precursor, with IRR CIs that did not overlap. For example, having a first symptom of depression compared with other possible first symptoms was not significantly associated with subsequent elated mood: IRR = 1.01 (95% CI: 0.65, 1.57), adjusted P = .98; in contrast, a first symptom of elated mood compared with other first symptoms was associated with higher likelihood of subsequent depression: IRR = 4.01 (95% CI: 2.41, 6.67), adjusted P < .001. As a first symptom, depression was associated with lower likelihood of subsequent subthreshold psychotic symptoms: IRR = 0.71 (95% CI: 0.56, 0.90), adjusted P = .03; but a first subthreshold psychotic symptom was associated with higher likelihood of subsequent depression: IRR = 2.36 (95% CI: 1.63, 3.41), adjusted P < .001.

Fig. 3.

Temporal associations between depression and other precursor symptoms. Modified Poisson regression with depression either as the predictor variable (ie, the first symptom; represented with a triangle) or the outcome variable (ie, a subsequent symptom; represented with a circle). Each line in the figure indicates temporal associations between depression and another symptom. Associations are not shown for pairs with <5 observations. Asterisks indicate P values <.05 after adjustment for false discovery rate. Note: NPS, nonpsychotic symptom; STPS, subthreshold psychotic symptom.

With suspiciousness (figure 4), we found nonoverlapping IRR CIs in 3 pairs: those involving anxiety, depression, and subthreshold psychotic symptoms in association with suspiciousness. For example, when the first symptom was suspiciousness, the IRR for subsequent anxiety was 3.40 (95% CI: 1.79, 6.46), adjusted P value = .002; but when the first symptom was anxiety, the IRR for subsequent suspiciousness was 1.15 (95% CI: 0.77, 1.73), adjusted P = .59. When the first symptom was suspiciousness, the IRR for another subsequent subthreshold psychotic symptom was 2.38 (95% CI: 1.25, 4.52), adjusted P = .03, but when the first symptom was any subthreshold psychotic symptom other than suspiciousness, the IRR for subsequent suspiciousness was 0.56 (95% CI: 0.32, 1.00), adjusted P = .09.

Fig. 4.

Temporal associations between suspiciousness and other precursor symptoms. Modified Poisson regression with suspiciousness either as the predictor variable (ie, the first symptom; represented with a triangle) or the outcome variable (ie, a subsequent symptom; represented with a circle). Each line in the figure indicates temporal associations between suspiciousness and another symptom. Associations are not shown for pairs with <5 observations. Asterisks indicate P values <.05 after adjustment for false discovery rate. Note: NPS, nonpsychotic symptom; STPS, subthreshold psychotic symptom.

Interactions of Temporal Associations With Sex and Diagnosis

Interactions with sex and diagnosis (affective and nonaffective psychosis) were not significant, except for 2 interactions with sex involving irritability as a subsequent symptom (supplementary table S2). After an initial subthreshold psychotic symptom relative to other possible first symptoms, there was a higher likelihood of subsequent irritability in male (IRR = 2.49; 95% CI: 1.57, 3.96) but not in female (IRR = 0.95; 95% CI: 0.50, 1.82) individuals (adjusted interaction P = .02). Conversely, after an initial nonpsychotic symptom, there was a lower likelihood of subsequent irritability in male (IRR = 0.40; 95% CI: 0.25, 0.64) but not in female (IRR = 1.05; 95% CI: 0.55, 2.00) individuals (adjusted interaction P = .02).

Discussion

In this systematic, catchment-based retrospective study of FEP, we found that combinations of precursor symptoms most commonly included nonpsychotic symptoms, particularly depression, anxiety, and substance use. This result is consistent with previous evidence that those symptoms (and their related diagnostic categories) were among the most prevalent in retrospective studies of the precursor stage of FEP^2,3 and prospective studies of the clinical high-risk state for psychosis.^33,34 The diverse admixtures of precursor symptoms can be further understood by extending their investigation not just across syndromes but also over time—with implications for early identification paradigms and drawing on recent staging models of mental illness.

Foundational research has identified common trajectories of the precursor stage before FEP, establishing that individuals tend to develop nonpsychotic symptoms before subthreshold psychotic ones, and the similarity of prodromal psychosis and depression until the point that psychotic symptoms appear.^3,4,10,12 Additional key findings are the prominence of negative symptoms,³⁵ and that depression and anxiety predict the persistence of paranoia in individuals at clinical high risk for psychosis.³⁴ The present study adds to these findings, and that of our own work,^14,18,22 by systematically examining temporal associations based on the initial precursor symptom, allowing a granular (symptom-specific) investigation of trajectories before FEP. Temporal associations varied in their magnitude and direction, with initial subthreshold psychotic symptoms (vs nonpsychotic symptoms), as well as specific first symptoms (suspiciousness, impaired role functioning), associated with higher likelihood of many subsequent symptoms. Identifying these trajectory features may have implications for risk prediction, service organization, and understanding mechanisms of illness progression.

Longitudinal trajectories of symptoms preceding FEP may provide useful data for risk prediction, such as if specific pairs or sequences of symptoms are differentially associated with progression to psychosis. As indicators of illness progression, the temporal sequences of symptoms experienced could be especially informative for clinical staging models, which aim to guide the tailoring of mental health services based on illness evolution over time.^36,37 Of the pairs explored here, those involving suspiciousness stand out: a first symptom of suspiciousness seemed to be a risk factor for many other subsequent symptoms, but the reciprocal associations of other symptoms preceding suspiciousness were not of the same magnitude. Suspiciousness was also an important predictor of psychosis conversion in NAPLS-1 and NAPLS-2, two cohorts of individuals at clinical high risk for psychosis.^38,39 Furthermore, in a group of homeless individuals with a history of psychotic disorder, suspiciousness seemed to precede the worsening of other psychotic symptoms.⁴⁰ We also found that a first symptom of impaired role functioning was associated with higher likelihood of many subsequent symptoms, in line with the observation that functional impairment before psychosis onset is associated with symptom severity during FEP.⁴¹ In NAPLS-1, impaired social functioning in early adolescence, but not academic or total functioning, was associated with psychosis conversion in clinical high-risk individuals.³⁹ Interestingly, this association with psychosis conversion was mediated by the presence of other prodromal symptoms including social anhedonia, echoing our result that a first symptom of impaired role functioning was associated with higher likelihood of subsequent social withdrawal and depression before FEP.

Together with the asymmetrical associations identified here, these findings suggest that risk calculators may be further optimized by integrating knowledge regarding the timing of onset and ordering of key symptoms such as suspiciousness, impaired role functioning, and other subthreshold psychotic and nonpsychotic symptoms. Importantly, temporal associations between precursor symptoms may reflect psychological and neurobiological mechanisms of illness progression. For example, the experience of suspiciousness—where the environment is interpreted as hostile or threatening—could lead to secondary depression or anxiety, eventually culminating into a psychotic episode that brings the person to mental health care. Hypothetically, the first onset of precursor symptoms such as suspiciousness or other subthreshold psychotic symptoms might also mirror critical phases in the etiopathogenesis of psychosis, such as exposure to socioenvironmental adversity,⁴² abnormal pruning and myelination,⁴³ striatal dopamine dysfunction,⁴⁴ or cortical glutamate dysfunction.⁴⁵ These environmental and biological mechanisms of illness evolution may be reciprocally compounded by the effects on physiology of other symptoms such as affective distress or substance use.

To date, the dominant approach to identifying individuals at risk of psychosis has been the detection of attenuated forms of psychotic illness, primarily defined according to the presence of subthreshold psychotic symptoms.⁴⁶ Yet given the diversity of clinical trajectories preceding the onset of psychosis,^47,48 a transsyndromic approach may also advance our understanding of the precursor stage: by measuring the presence or absence of a range of possible precursor symptoms, irrespective of their psychotic or nonpsychotic domains, we empirically determined their combinations prior to FEP and ranked them according to their prevalence. Of relevance from a risk identification perspective, the most prevalent combinations of symptoms varied by sex and diagnosis, in keeping with known sex differences in precursor symptoms (eg, higher prevalence of substance use and social functioning deficits in male and nonaffective psychosis).^3,49,50 If these symptom combinations are differentially associated with subsequent clinical needs and trajectories, they could contribute to risk assessments within early intervention services.

Contemporary clinical staging models have emphasized the pluripotential nature of many early-stage symptoms and syndromes, including but not limited to subthreshold psychotic symptoms.^37,51 Indeed, many of the combinations found in the present analysis are likely to be nonspecific predictors of various mental health problems—indexing not so much a single diagnostic risk as a transdiagnostic proneness to general psychopathology. However, other combinations, while still transdiagnostic, may be more specific for psychosis: combinations that are particularly predictive of subsequent FEP may be the co-occurrence of suspiciousness with other symptoms indexing affective distress (eg, depression, anxiety), over and above isolated suspiciousness or isolated affective symptoms, or a decrease in role functioning that seems to precede the onset of other nonpsychotic and subthreshold psychotic symptoms. In future research, the predictive validity of transdiagnostic symptom combinations during this early stage of illness could be examined prospectively, ideally in transdiagnostic youth services where different trajectories of illness (leading to both psychotic and nonpsychotic disorders) could be identified and compared.^47,52

Data regarding the order and sequences of symptoms prior to FEP can inform or validate theories about how symptoms influence each other over time. This symptom-based approach would have the ultimate aim of enabling intervention at specific points along the causal chains that uniquely underlie a person’s illness progression.¹⁶ For example, more intensive interventions could be offered for those with particular initial symptoms (eg, suspiciousness) or with the highest risk combinations of suspiciousness followed by depression or brief delusions. In keeping with the principles of transdiagnostic staging models, then, this symptom-based approach would address existing mental health problems while aiming to prevent (or reduce the risk of) new ones.

Strengths and Limitations

Participants in this study were recruited systematically in a catchment-based early intervention service, providing us with a large sample of individuals with diverse symptom trajectories prior to FEP. Notably, this includes individuals who did not endorse subthreshold psychotic symptoms in the precursor stage and who therefore may not have fulfilled criteria for the clinical high-risk syndrome for psychosis prior to their FEP onset.¹⁷ Characteristics of the analytic sample were similar to those of individuals excluded from analyses due to missing data, suggesting that the results are likely generalizable across the consenting population. Results may not generalize to other groups, however, including individuals not eligible for the FEP program (eg, those diagnosed with a psychosis that was solely induced by substances) or who did not consent to the research.

Retrospective study designs are limited by the potential for recall bias—as a function of educational attainment,⁴ insight, time elapsed since the first symptom, or social desirability. Considering that symptoms often emerge in clusters, isolating a single first symptom may also be prone to measurement error. To the extent possible, these challenges were mitigated by combining multiple sources of information, such as collateral history with relatives, a review of health and social records, and by using interview strategies such as anchor points to increase reliability and reproducibility. While data gathering (including interviews with patients) and the consensus-based meetings chaired by a senior psychiatrist focused on individual symptoms and attempted to distinguish between clinical and colloquial definitions of psychosis, another limitation is that the order of symptoms was only able to be characterized in terms of first and subsequent symptoms. We therefore could not evaluate temporal associations within subsequent symptoms. Lastly, due to limited observations for each sequence of symptoms, we employed a parsimonious statistical approach that assumed constant incidence rates over time; however, there may be time-varying associations, eg, if depression is more likely to emerge in the days rather than months following the onset of suspiciousness, or vice versa. Such dynamic patterns may differentially map onto the mechanisms of illness progression and subsequent outcomes. The small numbers of observations also constrained statistical power to examine less frequent temporal associations and their differences according to sex or diagnosis of affective vs nonaffective psychosis.

Future research has the potential to pursue the empirical characterization of the precursor stage using other sources of information, including prospective ones. Studies of electronic medical records⁵³ and national registries⁵⁴ have identified self-harm as a significant predictor of incident psychosis. In individuals with psychotic disorders, digital trace data could provide valuable information about previous precursor symptoms and their subsequent trajectories.⁵⁵ These data sources could complement neurobiological and exposure-based markers of psychosis risk. Ultimately, if such research finds the timing and sequence of precursor symptoms to be important for selecting preventative interventions, then it may shape the structure and function of early intervention services—such as by supporting the shift from diagnostically siloed to transsyndromic models of care.

Supplementary Material

Supplementary material is available at https://academic.oup.com/schizophreniabulletin/.

Funding

This work was supported by grants 68961 and MCT-94189 from the Canadian Institutes of Health Research, salary awards from the Fonds de Recherche du Quebec-Sante (R.J. and J.L.S.), the Canadian Institutes of Health Research (S.N.I.), and the Canada Research Chairs Program (A.K.M.), the Sackler Foundation, and the Lobeer Foundation.

Author Contribution

V.P. and J.L.S. designed the manuscript. V.P. conducted the analyses and wrote the initial draft. J.L.S. served as lead for data curation, funding acquisition, and supervision and served in a supporting role for conceptualization, methodology, and writing—review and editing. A.K.M., S.N.I., M.L., R.J., and J.L.S. were involved with the curation of the PEPP-Montreal database (data acquisition). All authors contributed to data interpretation and revised the work for important intellectual content.