Reductions in the N1 and P2 Auditory Event-Related Potentials in First-Hospitalized and Chronic Schizophrenia

Introduction

The auditory evoked potential (AEP) includes short-latency (brain stem), mid-latency (thalamic and cortical), and long-latency (cortical) responses. The long-latency potentials of the AEP comprise the N1 (N100) and P2 (P200) peaks, the so-called vertex potentials. These potentials are known to be tightly coupled to stimulus parameters; eg, their latency is decreased and their amplitude increased as sound intensity is increased. This has led to their classification as exogenous or sensory potentials. The vertex potentials do not represent the first volley of sensory information into the primary auditory cortex, which occurs on the order of tens of milliseconds rather than hundreds. In addition, the vertex potentials represent several components with different brain generators,¹ some of which are not purely sensory but are modulated by subjective mental operations. For example, attention to the tones can lead to an increase in the amplitudes of N1 and P2 and hence there is some endogenous aspect to the vertex potentials, eg, O'Donnell et al.² Thus, the N1 and P2 are probably best considered event-related potentials (ERPs) encompassing both exogenous and endogenous potentials.

Much research has examined the N1 to various auditory stimulus configurations in schizophrenia. The N1 elicited to standard tones on a relatively slow oddball ERP task (standard and target tones presented every 1–2 s) is reduced in chronic schizophrenia.^3–5 This reduction is present regardless of whether the subjects attended to all stimuli and counted the rare tones or ignored the tones and read a book² or simply did a passive task⁶ and so cannot reflect reduction in only the endogenous component of N1. Kayser et al⁷ reported reduced N1 to both complex tones and phonemes in schizophrenia. Using binaural single-tone AEP tasks, Saletu et al⁸ showed that N1 was reduced in all schizophrenia patients whether characterized by thought disorder or more negative symptoms. Kessler and Steinberg,⁹ however, argued that N1 reductions were present in paranoid but not undifferentiated or residual schizophrenia, although their samples were relatively small (9 residual/undifferentiated patients vs 9 controls). Adler et al¹⁰ showed reduced AEP N1 amplitudes at 1-s interstimulus intervals (ISIs) in schizophrenia. Connolly et al¹¹ showed reductions of N1 in schizophrenia using monaural stimuli. Boutros et al¹² reported reduced N1 amplitude (and increased latency) in schizophrenia to the first of 2 clicks presented in a gating task.

Javitt and colleagues focused on the effects of ISI on N1 amplitude. Shelly et al¹³ showed that schizophrenia patients showed increasing N1 abnormalities at longer ISIs. Javitt et al¹⁴ found in monkeys that N1 and P2 had different refractoriness periods for various ISIs as in humans¹⁵ and that phencyclidine (PCP) reduced the vertex potentials only at long ISIs. As suggested by Javitt, the N1 deficits may only occur in at long ISIs in schizophrenia, in line with the monkey PCP data above, reflecting deficit in a process that is selectively activated at long ISIs and relies on N-methyl-D-aspartate channel. Such a pattern may also reflect a ceiling effect at longer ISIs where controls continue to recruit nonrefractory processing neurons (the N1 gets quite large at longer ISIs), but schizophrenia patients have recruited all available resources due to reductions of auditory cortex gray matter. This differential ISI effect may explain why some studies see no reduction of N1 at shorter ISIs, eg, Umbricht et al,¹⁶ see Javitt.¹⁷ Paradoxically, using extremely long ISIs (>12 s), Roth et al¹⁸ reported reduced N1 in medicated schizophrenia patients but not in unmedicated patients. Clearly, there have been relatively few examinations of ISI effects on N1 in schizophrenia, and the results have been contradictory.

The issue of whether medication, rather than disease pathology, causes N1 reductions has been the subject of many investigations. The results have been equivocal. Pfefferbaum et al¹⁹ showed no reduced N1 in off-medication patients in an oddball task, but Ogura et al³ showed reduced N1 in off-medication patients. Connolly et al¹¹ showed smaller “N120” in unmedicated patients (minimum 6-wk washout) using dichotic stimuli. Within single-tone auditory evoked potential (AEP) tasks, Buchsbaum's review²⁰ indicated that medications did further reduce N1, though also noting negative findings. Adler et al¹⁰ reported no differences in N1 between medicated and unmedicated schizophrenia. Roemer and Shagass²¹ claimed medication reduced visual but not overall auditory N1 amplitudes. Josiassen et al²² tested 72 unmedicated schizophrenia subjects and 72 controls and reported smaller N1 to clicks in schizophrenia. Kessler and Steinberg⁹ reported reduced N1 in unmedicated paranoid schizophrenia but not in unmedicated residual/undifferentiated schizophrenia. Thus, the weight of the evidence suggests that N1 is likely unaffected directly by medication but perhaps somewhat by psychiatric status although reduced even in unmedicated patients.

There has been renewed interest in N1 in schizophrenia, largely driven by 2 findings. One is its possible role as an electrophysiological endophenotype. Several recent genetic studies have tested N1 amplitudes in twins concordant and discordant for schizophrenia and in unaffected family members. Ahveninen et al²³ showed that N1 was reduced in unaffected twins and was more marked in monozygotic twins than in dizygotic twins. Force et al²⁴ showed that relatives of schizophrenia patients were impaired in overall N1 amplitude but not in the general augmentation of N1 with attention. The second finding is that N1 is reduced in schizophrenia during hallucinations²⁵ and has been hypothesized recently to serve as a physiological sign of impaired corollary discharge during speech production^26,27 with such impairment having a causal role in hallucinations.

The P2 response has received much less attention. O'Donnell et al² reported smaller amplitude and shorter latency P2 in schizophrenia than in controls, which was hypothesized to reflect the absence of a component termed “P2b.” This component has a longer latency than sensory P2, and thus, the superposition of the 2 components in the presence of attention leads to a larger and later P2. Although present in controls, P2b was lacking in patients. Boutros et al¹² reported smaller P2 amplitude in schizophrenia to the first of 2 clicks presented in a gating task. O'Donnell et al⁵ reported reduced P2 in chronic schizophrenia but not in bipolar disorder. Karoumi et al²⁸ reported no reductions in P2 amplitude in chronically ill schizophrenia (ChSz) and their well siblings vs controls, although patients and siblings were generally one-third to one-half SD smaller than controls. Siblings showed longer peak latency P2 at Cz only.

If N1 and P2 are reduced in schizophrenia and are associated with some of the genetic risk factors for schizophrenia as suggested by the extant literature, then it is important for the ERPs to be reduced at the first hospitalization for schizophrenia rather than to develop with disease course. Only 2 articles report on N1 in schizophrenia at first episode, with conflicting results, and only one of those articles reported P2. Brown et al²⁹ used a relatively specific method of averaging N1 to standards (50-ms pips, 1.3-s ISI, 60-dB SPL) preceding vs following a target stimulus on an auditory oddball task in 40 first-episode schizophrenia patients, 40 patients chronically ill with schizophrenia, and matched controls. They found reduced N1 to standards before and after targets (and to targets) in first-episode schizophrenia, just like in chronically ill patients. By contrast, however, Valkonen-Korhonen et al³⁰ reported no N1 reductions in 25 first psychosis subjects to monaural (right ear) stimuli on an auditory oddball task (pips duration not specified, 1.0-s ISI, 55-dB nHL). Inspection of figure 1 from that article shows larger N1 in first-hospitalized psychosis patients than in matched controls. Caveats to the study include the fact that the psychotic patients studied were diagnostically mixed, with only 13/25 subjects meeting criteria for schizophrenia or schizophreniform disorder, and inclusion of unipolar (3) and bipolar (1) affective disordered subjects, and the fact that N1 amplitude was collapsed across 58 sites for analysis. Brown et al²⁹ reported statistically significant smaller P2 in first-hospitalized schizophrenia (FHSz) to standards after targets. The P2 to standards before targets was also smaller than in controls, but differences did not survive Bonferroni correction. To summarize, the evidence on N1 and P2 reductions in FHSz is scant and equivocal. Further, using the same number of trials to average standard and target ERPs, as in Brown et al,²⁹ is essential when comparing responses to standard and target stimuli to equate the signal-to-noise ratio (33 trials maximum in that study). However, averaging all available standards, as in Valkonen-Korhonen et al,³⁰ gives the best estimate of the true ERP with the least noise. It is not essential to examine target effects to determine if N1 and P2 are reduced in FHSz, and using all standard trials will provide a more accurate estimate of the true ERPs.

The present study examined whether N1 and P2 deficits were present in patients first hospitalized, or within 1 year of first hospitalization, for schizophrenia or schizophreniform disorders vs age-, parental socioeconomic status–, handedness-, gender-, and Wechsler Adult Intelligence Scale—Revised/Third Edition information subtest scaled score–matched controls. In addition, chronic schizophrenia patients and appropriately matched controls were tested to establish the pattern of deficits in clearly documented psychiatrically ill subjects. Two different paradigms were performed on 2 separate groups of samples as well, resulting in 2 groups of FHSz subjects and 2 groups of younger controls and 2 groups of chronically ill patients and 2 groups of older controls. Although examination of task parameter effects on N1 and P2 was not the main reason for the different paradigms (see “Stimuli” below), 2 separate samples allow for immediate replication of findings and comparison of effects across different stimulus configurations. For N1 and P2 to serve as powerful endophenotypes, they must be obligatorily reduced in schizophrenia, even at the time of first hospitalization rather than a consequence of illness duration. This is the main measure of interest in this article.

Methods

Results

ERP waveforms are presented in figure 1. Robust N1 and P2 potentials are evident.

Discussion

The results show that N1 and P2 are reduced in schizophrenia, both in chronically ill patients and at the first hospitalization. The reductions in N1 occurred with no latency differences between groups and likely reflect abnormalities of the primary N1 generators in superior temporal gyrus auditory cortices. The finding of shorter P2 peak latencies and smaller amplitudes in ChSz and FHSz compared with their matched controls replicates the pattern observed by O'Donnell et al^2,5 in chronically ill patients and are consistent with the absence of an endogenous, attention-related P2 component termed P2b. In well controls, this slightly later component overlaps the slightly earlier, more sensory P2 component(s), leading to a larger and later peak P2 amplitude.² Because all schizophrenia samples in this study showed earlier P2 latencies and smaller amplitudes, we speculate that the reduction in P2 relates to reductions in P2b, a component responsible for P2 augmentation during attention to tones within a specific sensory channel. The results in first-episode schizophrenia are consistent with findings of Brown et al²⁹ of reduced N1 and P2 to subsets of standards in FHSz. N1 and P2 reductions appear to be robust phenomena associated with schizophrenia. Their presence at first hospitalization suggest they may be suitable candidates for intermediate endophenotypes.

P2 is simply understudied in schizophrenia. We found reduced P2 with earlier peaks in all patient samples, consistent with the topography O'Donnell et al² attributed to deficient P2b. The finding is in contrast, however, with the observation of later P2 latencies in relatives of schizophrenia patients.²⁸ Definitive studies of N1 and P2 in relatives are necessary. If relatives do not show the same deficits observed in patients, then the utility of these ERPs as endophenotypes is questionable. Such studies are not only underway in several laboratories, but the N1 and P2 to frequent stimuli from the extant family studies that focused only on ERPs to the target stimuli (eg, P3) provide a large extant database for analysis that should be examined.

The lack of stable correlations with symptoms across both samples suggests that N1 and P2 amplitudes are little affected by the state of the subjects. Although N1 has been shown to be smaller during hallucinations, it does not appear to be smaller per se at all times in patients that hallucinate more than others. This in some way strengthens the argument for N1 and P2 as endophenotypes, as they seem more associated with the trait of being schizophrenic rather than the state of present symptom exacerbation.

Vertex potential reductions do not need to be pathognomonic for schizophrenia to be useful as endophenotypes. An endophenotype may well be expressed in the population but at relatively low frequency. Neither do all persons with schizophrenia need show a small N1 or small P2. Those who do, however, may have a certain genetic makeup as part of their entire constellation of genetic risk factors. N1 and P2 may serve as tools sensitive to the genetic risk factors. Other disorders may show N1 and P2 reductions. However, the path to such reductions is likely to be different in these diseases. Because N1 and P2, whatever their functional roles, are clearly in the information processing stream that is affected in schizophrenia, they are not simply epiphenomenal endophenotypes (as might be fingerprint whorls or hair swirls, eg) but rather are close to the functional pathway of impaired cognition in the disorder (namely, auditory processing, language, and thought).

There were moderate correlations between the dosage of medication and both N1 slowing and amplitude reduction. The effects sizes were rather small. There were no effects of medication classes in these relatively large samples. Thus, it seems unlikely that medications cause the initial N1 and P2 reductions, although medications appear to exert some modest effects on their timing and size. The following analysis was done to try to remove medication effects from the patient's Cz N1 amplitudes. Because no control took antipsychotics and medication overlapped entirely with diagnosis, analysis of covariance or multivariate analysis of covariance cannot be used. Instead, the unstandardized B weight for medication dose within patients, after regressing for chronicity and stimuli, was used to adjust N1 amplitude according to the formula: adjusted_Cz = raw_Cz − (.002 × meds). Although this method has many assumptions and should be interpreted cautiously, N1 differences between groups remained (P = .026), as did the main effect of stimulus (P = .002). Although reassuring, it remains unclear what role medication levels play in AEP reductions or whether the amount of medications deemed clinically necessary and the degree of N1 and P2 abnormalities are consequences of some other primary abnormality.

Although the general pattern on N1 and P2b reductions were the same in ChSz and FHSz, there were some subtle differences. The effect size for difference with matched controls was larger in ChSz than in FHSz, although substantial subject Ns would be necessary to attain statistical significance for any chronicity-related effect. The absence of background noise, leading to a perceptually louder tone, led to less P2b augmentation in FHSz than in ChSz. Given that the ChSz sample sizes were relatively small for the task without noise, caution should be used in assessing this complex interaction. It is likely that both groups showed moderate N1 and P2 augmentation to the louder stimuli.

One interesting finding in this study was the difference in P2 topography between younger and older subjects. Younger subjects, regardless of whether they were psychiatrically ill or well, showed markedly reduced frontal amplitude of P2. It is interesting to note that the age ranges between the “younger” and “older” samples were not that large. This effect seems to be a rather rapid change in the late 20s. This aging effect deserves further study to determine the underlying physiology and relation to brain maturation during the late third decade.

Several caveats need to be considered with regard to the data. We tested relatively few women, all in the first-hospitalized samples and younger controls. Thus, the degree to which these results generalize to women is not entirely clear. There are many factors that affect N1 and P2 amplitudes and latencies. We suggest most studies show abnormalities of vertex potentials in schizophrenia, but certainly some do not. Thus, the results in this study of abnormalities in the N1 and P2 elicited by standards in an oddball task do not mean that N1 and P2 in all tasks will be similarly reduced. It remains important to examine the N1 and P2 elicited by single tones with and without attention to determine the degree to which “sensory” components of N1 and P2 are affected early in the disease. For example, based on the topography of P2b described by O'Donnell et al², we infer P2b abnormalities may be present in first-hospitalized patients. We cannot, however address purely sensory aspects of P2 using the current task and analysis. Our previous work showed no reduction of MMN in a subset of these patients at first hospitalization (Salisbury et al 2004) and that MMN reduction developed during the early course of the disease³⁷. By contrast, P3 was reduced at first hospitalization³⁸. The vertex potentials here were reduced at first hospitalization. Although N1 and P2 are typically referred to as sensory potentials, they occur much later than the initial afferent signals to the auditory cortex and were measured from an attention-dependent task. It may be that the ERPs (N1, P2, MMN, P3) are sensitive to different underlying pathologies in different generator sites with different time courses. On the other hand, it may be that the MMN, generally thought to be preattentive, may not be as affected at first hospitalization as attention-related potentials like the N1 and P2 here and P3. First hospitalization likely postdates the actual onset of symptoms and may in fact miss the actual first psychotic episode. Although determination of prodrome onset would be optimal, we did not have reliable methods or the infrastructure necessary for such a determination during the time this data were collected. The standard method in the literature for reliable determination of onset is when symptoms are severe enough to require hospitalization, which has the advantage of being extremely objective.

In summary, N1 and P2b are reduced in FHSz patients, similarly to chronically ill patients, on an auditory oddball target detection ERP task with relatively slow presentation rates (1.2-s ISI). These classic vertex potentials may thus serve as potential endophenotype that are easy to record. The presence of deficits in first-hospitalized patients is necessary but not sufficient for establishing N1 and P2 as endophenotypes and is only one of the necessary criteria. Further research examining these potentials in family members (crucial for demonstrating genetic links) and using single-tone AEP tasks (crucial for isolating “sensory” generators) is warranted. With regard to the underlying neurophysiology, research is currently underway to examine whether the reductions in N1 and P2b amplitude can be associated with differences in phase synchrony. This is of some importance, as determining whether this difference reflects structural cortical generator loss or functional local cortical circuit derangement has implications for understanding the pathophysiology of schizophrenia.

Funding

Department of Veterans Affairs (Merit Award, Schizophrenia Center Award, Middleton Award to R.W.M.); National Institute of Health (R01 MH 40799, R01 MH 052807, P50MH080272 to R.W.M.; R01 MH58704 to D.F.S.); Mental Illness and Neuroscience Discovery foundation (to R.W.M.); and The National Alliance for Research in Schizophrenia and Depression (to D.F.S.).