The Difficulty in Finding Relationships Between ERPs and Clinical Symptoms of Schizophrenia

Who hasn’t thrown up their hands in despair when no significant correlations can be found between event-related potential (ERP) component amplitudes and symptoms of schizophrenia? The amplitudes of the typical auditory ERP components—P50, N100, P200, P300—are reduced in people with schizophrenia and their first-degree relatives, in schizophrenia patients early in their illness, and even in people at clinical high risk for schizophrenia. That is, many ERP components seem to be stable trait markers of psychosis and psychosis vulnerability. In light of their sensitivity to a psychosis diagnosis, it is interesting and troubling that these ERP components tend to be relatively insensitive to the symptoms that define psychosis.

A brief history of ERPs in psychiatry may help us understand this conundrum. When EEG was first used to study mental illness more than 50 years ago, the hope was that EEG abnormalities could provide a diagnostic “laboratory test” and a treatment target. Studies using resting state EEG (or quantitative EEG [QEEG]) preceded the use of ERPs; interest in QEEG waned with the advent of computers of average transients and their ability to extract stimulus-locked ERPs from the EEG. Some of the earliest ERP studies were designed to elicit the various ERP components passively, ¹ to minimally tax motivation and higher order processes, both of which are compromised in some people with schizophrenia. Furthermore, auditory paradigms were preferred over visual to avoid problems associated with people closing their eyes, diminished alertness, and so on. However, passive, simple auditory ERP paradigms are limited in what they can reveal about psychotic symptoms, as they are not specifically designed to provoke symptoms or the mechanisms by which the symptoms are manifest. Also contributing to the lack of correlations in the older published literature is the emphasis placed on diagnosis rather than symptoms, with some studies comparing ERPs across diagnoses. ²

Although much of the ERP literature involves passive paradigms, many investigators studying the pathophysiology of schizophrenia have used simple, but active, continuous performance tasks, such as the auditory oddball task. Oddball tasks requiring a response to infrequent target stimuli randomly imbedded in a series of frequent standard stimuli elicit a target P300. ³ It reflects a voluntary shift of attention to a task-relevant stimulus that calls for a motor response (press a button) or updating of memory (running count of targets). Accordingly, we might expect that reductions of P300 in schizophrenia would be related to difficulties maintaining attention, staying on task, and even engagement in life—amotivation and apathy. While there is a handful of articles showing a relationship between P300 amplitude and poor social/occupational functioning,^4,5 negative symptoms,^4,6 and disorganization, ⁴ it is surprising there are not more, given the number of papers showing a reduction of P300 in schizophrenia. ⁷ Similarly, while there are several studies showing the sensitivity of P300 to positive symptoms,^8-10 they are surprisingly few in number. Again, this could reflect a lack of relationship, a failure to report disappointing findings, or a failure to test for relationships. In addition, we may be asking too much of P300; why should P300 amplitude vary with the positive symptoms of schizophrenia, such as hallucinations and delusions? It taps neither these symptoms nor the underlying mechanisms contributing to these symptoms.

However, some investigators use ERPs to directly assess a symptom of schizophrenia. A good example is the N400 elicited in semantic priming tasks to assess loose associations typical of thought disorder. Again, considering the large number of studies using this paradigm, relatively few find a relationship between N400 and thought disorder or the subsymptoms (ie, loose associations) that contribute to thought disorder, although correlations between N400 and positive symptoms ¹¹ and positive thought disorder ¹² have been reported.

Where does this leave us? There are a number of reasons we may have difficulty with finding relationships between symptoms and ERP components. Primary among them may be our poor assessment of symptoms: negative symptom ratings rely on the patients’ reporting of daily activities and the clinicians’ observations of their behavior; positive symptoms rely on the patients’ ability to describe the unfathomable experiences they are having and the clinicians’ ability to understand and interpret them. While this might suggest that negative symptom ratings would be more valid and thereby correlate better with ERP components, this does not seem to be borne out in the literature.

Also contributing to our difficulty in relating symptoms to ERP components may be poor mechanistic models of the psychotic experience. Consistent with the RDoC (Research Domain Criteria) framework, ¹³ we might benefit from relating ERP components to specific domains of function that underlie specific symptoms. But again, we will encounter the need to elicit those ERPs in paradigms and settings designed to evoke or mimic the symptom or the underlying pathology.

We must also consider the obvious decoupling of the symptoms from the neurobiology due to medications and comorbidites. ¹⁴ For example, it is quite possible that medications differentially alter the outward manifestation of an underlying pathology (eg, a psychotic symptom) and the pathology itself (eg, deficient neuroplasticity). Finally, we must also consider the possibility that the schizophrenia diagnosis breeds truer than the symptoms it comprises. ¹⁵