P300 as a clinical assay: rationale, evaluation, and findings

Abstract

Use of the P300 event-related brain potential (ERP) as a clinical assay is reviewed and assessed by comparing its distribution qualities with normative biomedical testing data from published studies. The coefficient of variation statistic was calculated for P300 data and a variety of clinical testing data. P300 amplitude and latency variability was found to be highly comparable and sometimes superior to routinely employed biomedical assays. These results are discussed in terms of how to control inter-group ERP variability and the application of normative P300 data in clinical settings.

Introduction

Neuroelectric techniques have provided some of the most direct evidence for assessing central nervous system (CNS) function. Although standardized electroencephalographic (EEG) and evoked potential (EP) measures have been used clinically for many years (Halliday, 1993), cognitive event-related brain potentials (ERPs) have not as yet been employed in a routine fashion, despite considerable research demonstrating systematic differences between normal and patient population. In particular, the P300 component of the ERP has been widely applied in the scientific study of age-related cognitive dysfunction, because it reflects attentional and memory processes. However, the usefulness of P300 has been limited, in part because the definition of normal component values and a standard method for their acquisition have not been established (Polich, 1998). A related issue involves the inherent variability observed for ERP measures in general and especially when they are applied clinically, although this issue has not been addressed quantitatively. Toward that end, the present paper briefly reviews the background of clinical ERP applications and then compares normative P300 measures with biomedical test data to ascertain whether these ERP data are comparable to established clinical assays.

P300 amplitude is thought to index brain activity that is ‘required in the maintenance of working memory’ when the mental model of the stimulus environment (i.e. the context) is updated (Donchin et al., 1986, p. 256). This context updating theory has its roots in Sokolov’s model of the orienting response, which has been postulated to result from a change in the organism’s neural representation of the stimulus (Polich, 1989). P300 amplitude is also proportional to the amount of attentional resources devoted to a given task (Wickens et al., 1983, Kramer and Strayer, 1988) and has been associated with superior memory performance (Fabiani et al., 1990, Johnson, 1995). P300 amplitude can therefore be viewed as a measure of CNS activity that reflects the processing of incoming information when it is incorporated into memory representations of the stimulus and the context in which the stimulus occurs. Variation in P300 amplitude is, therefore, assumed to reflect the degree or quality with which that information is processed.

P300 latency is considered to be a measure of stimulus classification speed (Kutas et al., 1977, Polich, 1986a) and is generally unrelated to response selection processes (McCarthy and Donchin, 1981, Pfefferbaum et al., 1986). It is therefore independent of behavioral reaction time (Duncan-Johnson, 1981, Verleger, 1997). Indeed, it is just these properties that make the P300 a valuable tool for assessing cognitive function: Because P300 latency is an index of the processing time required before response generation, it is a sensitive temporal measure of the neural activity underlying the processes of attention allocation and immediate memory. In addition, P300 latency is negatively correlated with mental function in normal subjects, with shorter latencies associated with superior cognitive performance (e.g. Polich et al., 1983, Polich et al., 1990b, Emmerson et al., 1990, Polich and Martin, 1992). The neuropsychological tests that are best correlated with P300 latency are those that assess how rapidly subjects can allocate and maintain attentional resources. This association is also supported by results indicating that P300 latency increases as cognitive capability decreases from dementing illness (Squires et al., 1979, Brown et al., 1982, Homberg et al., 1986, Polich et al., 1986, Polich et al., 1990a, O’Donnell et al., 1992). Thus, P300 latency is directly associated with cognitive capability in both normal and patient populations.

The initial suggestion that the P300 component might be a useful tool for investigating cognitive function came from studies of normal aging and dementia, since peak latency was found to be prolonged in individuals with dementing illness compared to similarly aged normal subjects (Goodin et al., 1978a, Goodin et al., 1978b). Moreover, P300 latency increases systematically as cognitive function becomes worse in dementing illness, even though component size is not directly associated with the degree of mental impairment (Polich et al., 1986, Ball et al., 1989). Associations between P300 latency and the level of cognitive function also have been reported in neurological disorders, in confusional states, and for post-traumatic syndromes (cf. Squires et al., 1979, Hansch et al., 1982, Goodin et al., 1983, Homberg et al., 1986, O’Donnell et al., 1987, O’Donnell et al., 1990, Newton et al., 1989, Polich et al., 1992). Taken together, these findings suggest that P300 may be clinically useful as an index of cognitive function, although its diagnostic utility is questionable (cf. Brown et al., 1982, Pfefferbaum et al., 1984, Polich et al., 1986).

Some reports have suggested that ERP measures may distinguish between subcortical (e.g. Huntington’s and Parkinson’s disease) and cortical (Alzheimer’s, cerebral vascular accident) dementias (Rosenberg et al., 1985, Goodin and Aminoff, 1986). Other studies have indicated that P300 latency can separate individuals with dementia from those with depression-associated pseudodementia (Brown et al., 1982, Patterson et al., 1988). Discrimination between patients with early Alzheimer’s disease from normal subjects has also been reported (Polich et al., 1990a, Holt et al., 1995). Reliable differences in P300 amplitude and latency have been found between Alzheimer patients and controls, implying that if appropriate task conditions are employed ERPs can be used to evaluate this brain disease (Polich and Pitzer, 1999). Furthermore, the P300 component has been used to study psychiatric disorders such as alcoholism, depression, and schizophrenia (e.g. Pritchard, 1986, Courchesne, 1990, McCarley et al., 1993, Begleiter and Porjesz, 1995, Bruder et al., 1995, Boutros et al., 1997).

Fig. 1 illustrates some of these findings and highlights the current problems in the use of P300 in a clinical context: Any brain disorder that affects the primary cognitive operations of attention allocation and immediate memory will influence P300 measures by reducing amplitude and/or increasing latency. It follows, therefore, that until more detailed descriptions of P300’s neuropsychological origins and meaning are developed, the clinical utility of this ERP component is necessarily restricted to a general measurement of ‘cognitive efficiency’, i.e. how well an individual’s CNS can process and incorporate incoming information. Nevertheless, an objective, comparatively easy, and relatively inexpensive tool for assessing cognitive efficiency can be highly useful in both clinical and applied testing. In particular, the P300 ERP can be an effective and neuroelectric measure for evaluating therapeutic strategies involving CNS medications — a consideration that will become much more important as the neurochemical underpinnings of P300 generation become clarified.

The most common task used to elicit the P3000 is the two-stimulus discrimination or ‘oddball’ paradigm (Picton, 1992, Polich, 1999). Application of this approach over repeated testing produces good test–re-test correlation coefficients for both amplitude (0.50–0.80) and latency (0.40–0.70) measures (cf. Polich, 1986b, Fabiani et al., 1987, Karniski and Blair, 1989, Segalowitz and Barnes, 1993). However, variation in the test/re-test coefficients can occur from differences in inter-trial interval that can affect P300 habituation rates (Polich, 1989, Ravden and Polich, 1999), so that a series of trial blocks (e.g. 4–6) spaced approximately 10–15 min apart may help to assess consistency of a patient’s ERP response. Additional control over measurement variability, especially when normal and clinical groups are compared involves consideration of external factors that contribute to P300 variance. These effects may occur spontaneously or be induced by environmental variables, both of which have been characterized as ‘biological determinants’ of P300 (Polich and Kok, 1995).

Table 1 presents a summary of these variables. Because ERP studies often use within-subject designs, most biologically based factors that affect the P300 component do not vary excessively within a 2–3-h testing session, and their influence is usually minimized by using appropriate counterbalancing procedures. However, when between-subjects designs are employed as is typically the case in clinical studies (e.g. patients vs. normals or treatment vs. control), biological factors become critical and in these circumstances it is important to develop methods to reduce inter-group variability and the probability of hypothesis-testing error. Hence, ensuring that subjects are assessed similarly with respect to temporal variables (circadian, ultradian, seasonal, etc.), bodily functions (temperature, recency of food intake, fatigue, etc.), and matching patients on constitutional factors (age, gender, handedness, etc.) will reduce extraneous variance and facilitate statistical comparisons between groups (Polich and Kok, 1995, Polich, 1998). As noted, the correlation between neuropsychological test performance and P300 latency indicates that individual latency variability is related to mental processing speed, with peak timing affected by cognitive operations specific to perceptual and attentional processing and not just general intellectual capability. Thus, consideration of neuropsychological differences should prove to be useful for investigating ERP components that reflect specific aspects of information processing (e.g. Orlebeke et al., 1989, Pelosi et al., 1992a, Pelosi et al., 1992b, Colet et al., 1993, Stelmack and Houlihan, 1994).

Taken together, the thrust of the clinical P300 studies strongly suggest that this ERP component offers significant promise as a utilitarian measure of cognitive function. Despite this possibility, however, a general consensus is that the inherent variability of ERP measures mitigates against the practical use of such procedures (cf. Goodin, 1990, Pfefferbaum et al., 1990). This viewpoint is not without merit, since appreciable variation is found across normative P300 aging studies (Polich, 1996). The question therefore arises whether neuroelectric inter-subject variability is substantially more pronounced because of what ERPs reflect — CNS processing of stimulus information — than other basic bodily functions. More specifically, are P300 measures any different than other biomedical measurements, when these are used for to characterize wellness or illness?