Psychological Medicine, 1991, 21, 867-879 Printed in Great Britain
Long-latency auditory event-related potentials in schizophrenia and in bipolar and unipolar affective disorder WALTER J. MUIR,1 DAVID M. ST. CLAIR AND DOUGLAS H. R. BLACKWOOD From the Department of Psychiatry, University of Edinburgh
Long-latency auditory event-related potentials were examined in 96 subjects with schizophrenia, 99 with bipolar affective disorder and 48 with major depressive (unipolar) disorder, and compared with 32 in-patient and 213 normal controls. The latency of the P3 component was significantly greater in the schizophrenic and bipolar subjects compared to other groups. The difference was stable with respect to clinical state at the time of testing and was not due to age differences or the effect of psychotropic medications. The results support the clinical distinction between bipolar and unipolar affective disorders, but also show that P3 change is not specific to schizophrenia and found in bipolar but not unipolar affective disorder. SYNOPSIS
INTRODUCTION The search for boundaries delimiting the functional psychotic disorders has proved to be an elusive quest and yet our major diagnostic systems are based on assumed differences between schizophrenia, bipolar disorder and unipolar depression. Two central issues exist - the separation of schizophrenia from the affective psychoses, and the subdivision of affective illness. Each has been addressed in the past by a variety of clinical, genetic and biological techniques. Schizophrenia and bipolar disorder
Kraepelin (1919) placed great emphasis on the separation of schizophrenia from manic depressive psychosis and this distinction has been a cornerstone of psychiatric classification. However, the search for a clear clinical divide has been unsuccessful. A discriminant function analysis of clinical symptoms failed to provide clear evidence for two separate syndromes (Kendell & Gourlay, 1970) and there is no evidence of discontinuity between the illnesses 1 Address for correspondence: Dr Walter J. Muir, Department of Psychiatry, Kennedy Tower, Royal Edinburgh Hospital, Morningside Park, Edinburgh EH 10 5HF.
using outcome measures (Kendell & Brockington, 1980). The evidence from familial studies supports the distinction between schizophrenia and affective disorders in general, but an important study by Tsuang et al. (1980) of over a thousand firstdegree relatives of psychotic and control probandi showed the same excess of schizophrenia in relatives of bipolar probandi as seen in relatives of schizophrenics. The incidence of bipolar illness among relatives was around the same in the two groups. Further, the incidence of affective disorder with psychotic symptoms is increased in the first-degree relatives of schizophrenics compared to matched controls (Tsuang et al. 1985). Evidence for anatomical changes in schizophrenia such as increased ventricular size on CAT scanning has been found also in bipolars, and is associated with similar clinical features in both groups (Pearlson & Veroff, 1981; Nasrallah et al. 1982; Pearlson et al. 1984, 1985). On the basis of such evidence Flor-Henry (1983) advocated the concept of a symptomatological continuum of psychosis, an idea that has been further developed by Crow (1986). Attempts have been made to find a biological marker that distinguishes schizophrenia from bipolar disorder and much attention was focused
867
868
W. J. Muir and others
on alterations in neurotransmitters, their metabolites and the enzymes involved in their biosynthesis, all of which have been proposed abnormal in the psychoses. An early finding of low platelet monoamine oxidase (MAO) activity in schizophrenia (Murphy & Wyatt, 1972) seemed promising but there are medication effects (Maj et al. 1984) and the change is also seen in bipolar illness (Murphy & Weiss, 1972). Studies on mono-hydroxy phenyl glycol (MHPG), a monoamine metabolite, have centred on affective cases rather than schizophrenia (Lake et al. 1987); the serotonin metabolite 5-hydroxyindole acetic acid (5HIAA) concentrations do not seem to be consistently abnormal in schizophrenia (Stahl & Wets, 1987); and plasma levels of the dopamine metabolite homovanillic acid (HVA) may be strongly related to the clinical severity of the illness (Davis et al. 1985). Some physiological measures have been found to be consistently abnormal in schizophrenia and eye tracking dysfunction has been one of the most extensively studied. Since Holzman's original work showing the deficit in schizophrenia (Holzman et al. 1973), it has been well established that in schizophrenia smooth pursuit eye movements are interrupted by saccadic tracking and intrusions (Levin et al. 1982) and this dysfunction is not due to lack of motivation or simple inattention (Lipton et al. 1980a, b; Iacono et al. 1981; Levin et al. 1982). The abnormality is stable with respect to both neuroleptic medication and remission of psychotic symptoms (Iacono et al. 1981; Levy et al. 1983). There is a much lower incidence of the abnormality in the first degree relatives of bipolar compared to schizophrenic probandi (Holzman et al. 1984), suggesting that its genetic transmission is specific to schizophrenia. In these studies however abnormality was defined in a way which included 8% of the general population (Holzman et al. 1973, 1974). Eye movement dysfunction is also found in bipolar subjects (Shagass et al. 1974) but recently it has been suggested that this could be an effect of lithium medication (Iacono et al. 1982; Levy et al. 1985). Long-latency event-related potential (ERP) studies have shown consistent changes in schizophrenia. The P3 waveform, elicited by sensory discrimination tasks to stimuli that are rare or
unexpected, has been extensively examined in schizophrenia and a reduced amplitude when compared to controls has been consistently noted using auditory, visual and somatosensory modalities (Roth & Cannon, 1972; Levit et al. 1973; Shagass et al. 1977; Verleger & Cohen, 1978; Pass et al. 1980; Roth et al. 1980, 1986; Baribeau-Brown et al. 1983; Brecher & Begleiter, 1983; Morstyn et al. 1983; Faux et al. 1988). However, the latency of P3 was not found to differentiate among the groups but in three more recent investigations of the auditory P3 it has been found to be significantly prolonged in schizophrenia and seems to be independent of medication effects and clinical state (Pffeferbaum et al. 1984; Blackwood et al. 1987; Romani et al. 1987). The abnormality is also present in borderline personality disorder but not in major depressive disorder or a variety of neuroses and non-borderline personality disorders (Blackwood et al. 1986; Kutcher et al. 1987). These features, along with the apparent high degree of heritability indicated in twin studies (Surwillo, 1980; Polich & Burns, 1987) suggest that P3 latency may be an important biological trait marker for schizophrenia and allied conditions. The subdivision of affective disorder The classification of affective illness into bipolar and monopolar (now unipolar) subtypes by Leonhard (1957) has served as a useful framework for research into these disorders. Over the last thirty years several lines of investigation have lent support to the distinction of these illnesses showing differences in the familial distribution, natural history, and response to treatment. The rate of illness found among first degree relatives differs significantly for bipolar and unipolar probandi, the risk of affective illness being greater in relatives of bipolars (Angst, 1966; Perris, 1966; Winokur et al. 1969). Twin studies have revealed a greater concordance between bipolar monozygotic pairs than between unipolar monozygotic pairs (Allen, 1976; Bertelsen et al. 1977). Differences have also been found between the natural histories of the subgroups. In a large multi-centre study of over a thousand cases Angst and his colleagues found a significantly lower mean age at onset of illness for bipolar cases. Bipolar patients also tended to show a
Long-latency auditory event-related potentials
greater number of illness episodes and to have a poorer overall social adjustment (Angst et al. 1973). Treatment studies have revealed further differences. Prophylactic tricyclic antidepressants seem to reduce the risk of relapse in unipolar but not bipolar illness (Prien et al. 1973; Medical Research Council Drug Trials Subcommittee, 1981). Recently, Silverstone (1985) has claimed greater efficacy for bromocriptine in bipolar disorder, although this finding awaits replication. Although the clinical and genetic evidence strongly supports the validity of two separate entities and the concept of a dichotomy within affective illness is embraced by both the Research Diagnostic Criteria (Spitzer et al. 1978) and the DSM-III-R (American Psychiatric Association, 1987) under the rubrics of 'bipolar disorder' and 'major depressive disorder' the search for biological correlates of affective illness has yielded varied results. Neuroendocrine and neurotransmitter studies have not provided clear means for dividing affective illness. Dexamethasone non-suppression has been claimed by some to be infrequent in bipolar disorder but common in unipolar disorder (Schatzberg et al. 1983). Others have not found this (Stokes et al. 1984) and overall the interpretation of the dexamethasone test remains uncertain (Berger et al. 1984; Braddock, 1986). Studies on monoamines, their precursors, and their metabolites have yielded a wealth of contradictory findings. For example, in recent studies of the monoamine metabolite MHPG in the cerebrospinal fluid of depressed subjects an increase (Vestergaard et al. 1978; Koslow et al. 1983), a decrease (Subrahmanyam, 1975), or no change in concentration (Berger et al. 1980) have all been reported. Similarly HVA has been found increased (Vestergaard et al. 1978), decreased (Van Praag et al. 1973), and unchanged (Berger et al. 1980). The confusion persists for the serotonin metabolite 5-HIAA with no clear distinction found between bipolar and unipolar disorder. ERPs have been used for some time in the study of affective disorder but the use of different nosologies by different workers has made the interpretation of the findings difficult. Bipolars differed from unipolars when tested using
869
an augmenting-reducing paradigm with visual stimuli (Buchsbaum et al. 1973). Later studies however indicated that the amplitude of event related potentials in depression is related to clinical state (Friedman & Meares, 1979; Blackwood et al. 1987). Shagass and his colleagues have accumulated extensive ERP data in several sensory modalities in psychiatric populations but have chosen to view depressive illness as a continuum and to group bipolar depressed along with other psychotic depressed cases (Shagass et al. 1978, 1979, 1980, 1982; Straumanis et al. 1982). Others have chosen to look at depression alone (Pffeferbaum et al. 1984), or to use the primary-secondary distinction (Diner et al. 1985). Purpose of the present study
Our own studies have shown that the latency of the auditory P3 ERP elicited by a two-tone discrimination task distinguished schizophrenia and borderline personality disorders from unipolar depression and from other neuroses and non-borderline personality disorders (Blackwood et al. 1987; Kutcher et al. 1987). The present study was designed to show whether P3 and other long latency auditory ERPs distinguished schizophrenics from bipolar disorder and whether there were any differences between bipolar disorder and major depressive disorder.
METHOD Subjects All subjects were in-patients or out-patients of the Royal Edinburgh Hospital. Demographic characteristics of the groups are listed in Table 1. All subjects were under 60 years of age. Our previous work (St. Clair & Blackwood, 1985; Blackwood et al. 1988) showed a marked increase in the latency of P3 after, but not before, this age. Patients with schizophrenia All patients were able to give informed consent for the study and were assessed using the Schedule for Affective Disorders and Schizophrenia - a lifetime version (SADS-L; Endicott & Spitzer, 1978) by one of the authors. Ninetysix met the RDC criteria for definite schizophrenia. Five other subjects fulfilled the criteria
870
W. J. Muir and others Table 1. Demographic characteristics of groups
Controls Schizophrenics Bipolars Major depressed Non-psychotic psychiatric disorder
N
Mean age (s.D.) yr
Age range yr
213 96 88 46 32
30-5(11-7) 31-5(11-0) 35-2 (120) 38-2(13-2) 30-6(11-7)
16-59 15-57 17-59 18-60 17-60
but were too ill to cooperate fully with the physiological testing.
Males
Females
123 67 54 21
90 29 34 25 21
Normal control group Two hundred and thirteen were volunteers from the local community recruited through general practitioners and some hospital staff and university students were also included. None were on psychotropic medications, and they gave no history of psychiatric or neurological illness.
et al. 1978; Sklare & Lynn, 1984; Polich et al. 1985). To enable us to examine large numbers of frequently disturbed psychiatric patients we chose to use a single recording electrode for simplicity. Bipolar recordings were made between Ag/AgCl electrodes at the Cz position and an indifferent electrode at the left ear-lobe. A ground electrode was positioned at the right ear-lobe. In all cases electrode impedances were less than 2 kQ. Recordings were made within a sound-attenuated screened room. Auditory stimuli were delivered binaurally through headphones at a level of 70 db and with an inter-stimulus interval of 1-1 s. Subjects were instructed to silently count randomly-presented tones of 1-5 kHz which occurred in a regular series of lowerpitched 1-0 kHz tones. The overall ratio of highpitched to low-pitched tones was 1:9. The EEG signal was amplified by 10000 and passed through analogue filters which were set from 1 to 30 Hz (3 db down). The signal was then digitized at a rate of 1 kHz over a time epoch of 750 ms. A total of 75 ms of EEG was captured prior to delivery of the auditory stimulus to serve as a baseline for subsequent measurements. A total of 500 epochs was recorded and the data for high and low pitched tones averaged separately. An artefact reject circuit excluded all trials where the voltage exceeded + 45 /iV. Two recordings were made with each subject.
Recording of auditory event-related potentials Event-related potentials were recorded using a Nicolet CA1000 clinical signal averaging computer. Recordings were carried out exactly as in our previous studies (Blackwood et al. 1987, 1988; Kutcher et al. 1987; St. Clair et al. 1989; Muir et al. 1988) and the method is identical to that used by as number of other groups (Goodin
ERP rating Investigators who were blind to the subjects' diagnoses independently measured latencies and baseline-to-peak amplitudes of the individual components of the recorded waveforms. As in our previous studies these were designated as N l , P2, N2 and P3. Nl was the point of maximum negativity in the region 70—120 ms post-stimulus and P2 was the maximum posi-
Patients with affective disorders These subjects were selected on the basis of a clinical diagnosis of bipolar disorder or major depressive disorder. Eighty-eight met RDC criteria for bipolar disorder, of whom 75 were bipolar-I and 13 bipolar-II. Nine subjects met the criteria for bipolar-I disorder but were too ill (largely manic) to cooperate. Forty-eight patients met RDC criteria for major depressive disorder. In-patients with non-psychotic psychiatric disorders Thirty-two in-patients with non-psychotic disorders were used as a comparison group. None met the criteria for borderline personality disorder using the Diagnostic Interview for Borderlines (DIB; Gunderson et al. 1981). Details of the diagnoses of this mixed group of disorders are given in Table 2.
Long-latency auditory event-related potentials
Table 2. Diagnoses of non-psychotic psychiatric disordered group RDC Diagnosis
N
Minor depressive disorder 9 Intermittent depressive disorder 3 Anorexia nervosa 9 (3 also with minor depression) Bulimia nervosa Obsessive compulsive disorder Generalized anxiety disorder Panic disorder Passive-dependent personality disorder Unspecified personality disorder Substance abuse disorder
tivity between 140 and 230 ms in the averaged data of the low-pitched frequent stimuli. P3 was defined as the maximum positivity between 260 and 500 ms and the N2 the maximum negativity immediately preceding the P3 wave in the averaged responses to the high-pitched rare tones. The baseline amplitude was determined from the prestimulus record in each case, and amplitude of the individual components was measured with respect to this from the penrecorded tracings. Latencies were measured from the stimulus to the point of maximum amplitude of each wave component using the cursor provided by the computer on the visual display unit and by inspection of the pen-recorded trace. When the deflection did not form a sharp peak the latency was measured at the point of intersection of the tangents to the upgoing and downgoing slopes measured from the penrecorded trace by the method of Goodin et al. (1978). The mean of the two separate recordings was taken for the final results. Separate investigators, blind to the ERP measures made all the diagnoses. Clinical state at time of testing The degree of depression at the time of testing was assessed using Hamilton rating scores (Hamilton, 1960) in 61 of the subjects with bipolar disorder. Three subgroups were defined: 14 subjects were in a depressive phase of the illness with Hamilton scores of 17 or over (bipolar depressed); 20 subjects had Hamilton scores of 10 or less and did not fulfil the RDC criteria based on SADS-L evaluation for current manic or hypomanic episode (bipolar non-
871
Table 3. Characteristics of clinical subgroups of the affective disorders (For details of classification see text.) N
Mean age (s.D.)
Female:Male
Bipolar subgroups Bipolar depressed Bipolar non-depressed Bipolar manic
14 20 24
36-4(130) 33-3(110) 39-5(13-3)
7:7 9:11 7:17
Unipolar subgroups Unipolar depressed Unipolar non-depressed
22 11
38-9 (12-4) 34-0(12-2)
14:8 5:6
depressed); thirdly a group of 24 subjects fulfilled the RDC criteria for current manic or hypomanic episode (bipolar manic). A further 3 subjects had Hamilton scores between 10 and 17. Similarly 39 subjects were diagnosed as having major depressive disorder (RDC criteria based on SADS-L evaluation) and two subgroups of these were defined on the basis of Hamilton rating scores. A score of 17 or over was used to define a subgroup of 22 subjects who were clinically depressed at the time of testing (unipolar depressed) and this compared with 11 cases who scored 10 or below (unipolar nondepressed). There were a further 6 subjects who had Hamilton scores between 10 and 17. Demographic data on these subgroups are given in Table 3. Statistical analysis All results were analysed using the Statistical Package for the Social Sciences (SPSS-X, 1986). The ERP data followed approximately normal distributions and patient groups were compared using one-way analysis of variance (ANOVA) and the significance of the difference between the means was tested using Scheffe's procedure. Where appropriate analysis of covariance was also performed using age as a covariate. Within group comparisons were evaluated using Student's t tests.
RESULTS The groups differed significantly in age, (F(4,470) = 5-86; P< 0-0001). Subjects with major depressive (P < 0-01) and bipolar disorders (P
Controls No significant differences Schizophrenia > Controls Bipolar disorder > Controls Schizophrenia > Non-psychotic disorders Bipolar disorder > Non-psychotic disorders Schizophrenia > Major depressive disorder Schizophrenia > Controls Bipolar disorder > Controls Schizophrenia > Major depressive disorder Bipolar disorder > Major depressive disorder Schizophrenia > Non-psychotic disorders Bipolar disorder > Non-psychotic disorders
P
Controls Schizophrenia > Non-psychotic disorders Bipolars > Controls Schizophrenia > Controls Schizophrenia > Major depressives Schizophrenia > Non-psychotic disorders Bipolars > Controls Bipolars > Major depressives
P3
P < 0001
P< 0-001 P < 0001 P < 0001
P < 005
Table 10 b. ERP amplitudes (/JV + S.D. (TV)) of drug-free subject by group (Comparisons by Scheffe's procedure following ANOVA.)
Controls Schizophrenia Major depressive disorder Bipolar disorder Non-psycholic disorders Nl P2 N2 P3
Nl
P2
3-4±l-8(2U) 3 0 + 1-9 (46) 3-5 ±3-5 (35) 3-6 + 2-8(27) 3-6+1-7(20)
4-0+1-7(210) 3-2 + 1-4(46) 3-8+ 1-8 (35) 3-2+1-6(27) 3-4 ±1-7 (20)
N2
P3
3-6 + 2-4 (169) 2-6+1-5(38) 31 ±2-8 (31) 4-2 + 2-7(22) 31 + 1-9(15)
9-8 + 2-9(212) 6-0 + 2-8(44) 8-0 + 303(35) 6-5 + 2-7(26) 9-1+3-5(20)
No significant differences No significant differences No significant differences Schizophrenia < Controls Bipolars < Controls Schizophrenia < Non-psychotic disorders Major Depressives < Controls
1987) and indicate a refractoriness of P3 latency change to the patient's medication status or current mood state. P3 latency is delayed in both schizophrenics and bipolars in contrast to unipolar affective disorder. This provides psychophysiological support to the clinical distinction between bipolar and unipolar forms of depression, a concept which is further strengthened by the latest family studies on affective disorder (Andreasen et al. 1987). There is no evidence to suggest a physiological dichotomy between bipolar I and II subtypes. The methods used in our study have the advantage of being simple, easy to use and well tolerated by patients showing a wide variety of clinical states including those who were very disturbed. However, it does limit the psychological and physiological interpretation of the
P< P< P< P
Long-latency auditory event-related potentials in schizophrenia and in bipolar and unipolar affective disorder WALTER J. MUIR,1 DAVID M. ST. CLAIR AND DOUGLAS H. R. BLACKWOOD From the Department of Psychiatry, University of Edinburgh
Long-latency auditory event-related potentials were examined in 96 subjects with schizophrenia, 99 with bipolar affective disorder and 48 with major depressive (unipolar) disorder, and compared with 32 in-patient and 213 normal controls. The latency of the P3 component was significantly greater in the schizophrenic and bipolar subjects compared to other groups. The difference was stable with respect to clinical state at the time of testing and was not due to age differences or the effect of psychotropic medications. The results support the clinical distinction between bipolar and unipolar affective disorders, but also show that P3 change is not specific to schizophrenia and found in bipolar but not unipolar affective disorder. SYNOPSIS
INTRODUCTION The search for boundaries delimiting the functional psychotic disorders has proved to be an elusive quest and yet our major diagnostic systems are based on assumed differences between schizophrenia, bipolar disorder and unipolar depression. Two central issues exist - the separation of schizophrenia from the affective psychoses, and the subdivision of affective illness. Each has been addressed in the past by a variety of clinical, genetic and biological techniques. Schizophrenia and bipolar disorder
Kraepelin (1919) placed great emphasis on the separation of schizophrenia from manic depressive psychosis and this distinction has been a cornerstone of psychiatric classification. However, the search for a clear clinical divide has been unsuccessful. A discriminant function analysis of clinical symptoms failed to provide clear evidence for two separate syndromes (Kendell & Gourlay, 1970) and there is no evidence of discontinuity between the illnesses 1 Address for correspondence: Dr Walter J. Muir, Department of Psychiatry, Kennedy Tower, Royal Edinburgh Hospital, Morningside Park, Edinburgh EH 10 5HF.
using outcome measures (Kendell & Brockington, 1980). The evidence from familial studies supports the distinction between schizophrenia and affective disorders in general, but an important study by Tsuang et al. (1980) of over a thousand firstdegree relatives of psychotic and control probandi showed the same excess of schizophrenia in relatives of bipolar probandi as seen in relatives of schizophrenics. The incidence of bipolar illness among relatives was around the same in the two groups. Further, the incidence of affective disorder with psychotic symptoms is increased in the first-degree relatives of schizophrenics compared to matched controls (Tsuang et al. 1985). Evidence for anatomical changes in schizophrenia such as increased ventricular size on CAT scanning has been found also in bipolars, and is associated with similar clinical features in both groups (Pearlson & Veroff, 1981; Nasrallah et al. 1982; Pearlson et al. 1984, 1985). On the basis of such evidence Flor-Henry (1983) advocated the concept of a symptomatological continuum of psychosis, an idea that has been further developed by Crow (1986). Attempts have been made to find a biological marker that distinguishes schizophrenia from bipolar disorder and much attention was focused
867
868
W. J. Muir and others
on alterations in neurotransmitters, their metabolites and the enzymes involved in their biosynthesis, all of which have been proposed abnormal in the psychoses. An early finding of low platelet monoamine oxidase (MAO) activity in schizophrenia (Murphy & Wyatt, 1972) seemed promising but there are medication effects (Maj et al. 1984) and the change is also seen in bipolar illness (Murphy & Weiss, 1972). Studies on mono-hydroxy phenyl glycol (MHPG), a monoamine metabolite, have centred on affective cases rather than schizophrenia (Lake et al. 1987); the serotonin metabolite 5-hydroxyindole acetic acid (5HIAA) concentrations do not seem to be consistently abnormal in schizophrenia (Stahl & Wets, 1987); and plasma levels of the dopamine metabolite homovanillic acid (HVA) may be strongly related to the clinical severity of the illness (Davis et al. 1985). Some physiological measures have been found to be consistently abnormal in schizophrenia and eye tracking dysfunction has been one of the most extensively studied. Since Holzman's original work showing the deficit in schizophrenia (Holzman et al. 1973), it has been well established that in schizophrenia smooth pursuit eye movements are interrupted by saccadic tracking and intrusions (Levin et al. 1982) and this dysfunction is not due to lack of motivation or simple inattention (Lipton et al. 1980a, b; Iacono et al. 1981; Levin et al. 1982). The abnormality is stable with respect to both neuroleptic medication and remission of psychotic symptoms (Iacono et al. 1981; Levy et al. 1983). There is a much lower incidence of the abnormality in the first degree relatives of bipolar compared to schizophrenic probandi (Holzman et al. 1984), suggesting that its genetic transmission is specific to schizophrenia. In these studies however abnormality was defined in a way which included 8% of the general population (Holzman et al. 1973, 1974). Eye movement dysfunction is also found in bipolar subjects (Shagass et al. 1974) but recently it has been suggested that this could be an effect of lithium medication (Iacono et al. 1982; Levy et al. 1985). Long-latency event-related potential (ERP) studies have shown consistent changes in schizophrenia. The P3 waveform, elicited by sensory discrimination tasks to stimuli that are rare or
unexpected, has been extensively examined in schizophrenia and a reduced amplitude when compared to controls has been consistently noted using auditory, visual and somatosensory modalities (Roth & Cannon, 1972; Levit et al. 1973; Shagass et al. 1977; Verleger & Cohen, 1978; Pass et al. 1980; Roth et al. 1980, 1986; Baribeau-Brown et al. 1983; Brecher & Begleiter, 1983; Morstyn et al. 1983; Faux et al. 1988). However, the latency of P3 was not found to differentiate among the groups but in three more recent investigations of the auditory P3 it has been found to be significantly prolonged in schizophrenia and seems to be independent of medication effects and clinical state (Pffeferbaum et al. 1984; Blackwood et al. 1987; Romani et al. 1987). The abnormality is also present in borderline personality disorder but not in major depressive disorder or a variety of neuroses and non-borderline personality disorders (Blackwood et al. 1986; Kutcher et al. 1987). These features, along with the apparent high degree of heritability indicated in twin studies (Surwillo, 1980; Polich & Burns, 1987) suggest that P3 latency may be an important biological trait marker for schizophrenia and allied conditions. The subdivision of affective disorder The classification of affective illness into bipolar and monopolar (now unipolar) subtypes by Leonhard (1957) has served as a useful framework for research into these disorders. Over the last thirty years several lines of investigation have lent support to the distinction of these illnesses showing differences in the familial distribution, natural history, and response to treatment. The rate of illness found among first degree relatives differs significantly for bipolar and unipolar probandi, the risk of affective illness being greater in relatives of bipolars (Angst, 1966; Perris, 1966; Winokur et al. 1969). Twin studies have revealed a greater concordance between bipolar monozygotic pairs than between unipolar monozygotic pairs (Allen, 1976; Bertelsen et al. 1977). Differences have also been found between the natural histories of the subgroups. In a large multi-centre study of over a thousand cases Angst and his colleagues found a significantly lower mean age at onset of illness for bipolar cases. Bipolar patients also tended to show a
Long-latency auditory event-related potentials
greater number of illness episodes and to have a poorer overall social adjustment (Angst et al. 1973). Treatment studies have revealed further differences. Prophylactic tricyclic antidepressants seem to reduce the risk of relapse in unipolar but not bipolar illness (Prien et al. 1973; Medical Research Council Drug Trials Subcommittee, 1981). Recently, Silverstone (1985) has claimed greater efficacy for bromocriptine in bipolar disorder, although this finding awaits replication. Although the clinical and genetic evidence strongly supports the validity of two separate entities and the concept of a dichotomy within affective illness is embraced by both the Research Diagnostic Criteria (Spitzer et al. 1978) and the DSM-III-R (American Psychiatric Association, 1987) under the rubrics of 'bipolar disorder' and 'major depressive disorder' the search for biological correlates of affective illness has yielded varied results. Neuroendocrine and neurotransmitter studies have not provided clear means for dividing affective illness. Dexamethasone non-suppression has been claimed by some to be infrequent in bipolar disorder but common in unipolar disorder (Schatzberg et al. 1983). Others have not found this (Stokes et al. 1984) and overall the interpretation of the dexamethasone test remains uncertain (Berger et al. 1984; Braddock, 1986). Studies on monoamines, their precursors, and their metabolites have yielded a wealth of contradictory findings. For example, in recent studies of the monoamine metabolite MHPG in the cerebrospinal fluid of depressed subjects an increase (Vestergaard et al. 1978; Koslow et al. 1983), a decrease (Subrahmanyam, 1975), or no change in concentration (Berger et al. 1980) have all been reported. Similarly HVA has been found increased (Vestergaard et al. 1978), decreased (Van Praag et al. 1973), and unchanged (Berger et al. 1980). The confusion persists for the serotonin metabolite 5-HIAA with no clear distinction found between bipolar and unipolar disorder. ERPs have been used for some time in the study of affective disorder but the use of different nosologies by different workers has made the interpretation of the findings difficult. Bipolars differed from unipolars when tested using
869
an augmenting-reducing paradigm with visual stimuli (Buchsbaum et al. 1973). Later studies however indicated that the amplitude of event related potentials in depression is related to clinical state (Friedman & Meares, 1979; Blackwood et al. 1987). Shagass and his colleagues have accumulated extensive ERP data in several sensory modalities in psychiatric populations but have chosen to view depressive illness as a continuum and to group bipolar depressed along with other psychotic depressed cases (Shagass et al. 1978, 1979, 1980, 1982; Straumanis et al. 1982). Others have chosen to look at depression alone (Pffeferbaum et al. 1984), or to use the primary-secondary distinction (Diner et al. 1985). Purpose of the present study
Our own studies have shown that the latency of the auditory P3 ERP elicited by a two-tone discrimination task distinguished schizophrenia and borderline personality disorders from unipolar depression and from other neuroses and non-borderline personality disorders (Blackwood et al. 1987; Kutcher et al. 1987). The present study was designed to show whether P3 and other long latency auditory ERPs distinguished schizophrenics from bipolar disorder and whether there were any differences between bipolar disorder and major depressive disorder.
METHOD Subjects All subjects were in-patients or out-patients of the Royal Edinburgh Hospital. Demographic characteristics of the groups are listed in Table 1. All subjects were under 60 years of age. Our previous work (St. Clair & Blackwood, 1985; Blackwood et al. 1988) showed a marked increase in the latency of P3 after, but not before, this age. Patients with schizophrenia All patients were able to give informed consent for the study and were assessed using the Schedule for Affective Disorders and Schizophrenia - a lifetime version (SADS-L; Endicott & Spitzer, 1978) by one of the authors. Ninetysix met the RDC criteria for definite schizophrenia. Five other subjects fulfilled the criteria
870
W. J. Muir and others Table 1. Demographic characteristics of groups
Controls Schizophrenics Bipolars Major depressed Non-psychotic psychiatric disorder
N
Mean age (s.D.) yr
Age range yr
213 96 88 46 32
30-5(11-7) 31-5(11-0) 35-2 (120) 38-2(13-2) 30-6(11-7)
16-59 15-57 17-59 18-60 17-60
but were too ill to cooperate fully with the physiological testing.
Males
Females
123 67 54 21
90 29 34 25 21
Normal control group Two hundred and thirteen were volunteers from the local community recruited through general practitioners and some hospital staff and university students were also included. None were on psychotropic medications, and they gave no history of psychiatric or neurological illness.
et al. 1978; Sklare & Lynn, 1984; Polich et al. 1985). To enable us to examine large numbers of frequently disturbed psychiatric patients we chose to use a single recording electrode for simplicity. Bipolar recordings were made between Ag/AgCl electrodes at the Cz position and an indifferent electrode at the left ear-lobe. A ground electrode was positioned at the right ear-lobe. In all cases electrode impedances were less than 2 kQ. Recordings were made within a sound-attenuated screened room. Auditory stimuli were delivered binaurally through headphones at a level of 70 db and with an inter-stimulus interval of 1-1 s. Subjects were instructed to silently count randomly-presented tones of 1-5 kHz which occurred in a regular series of lowerpitched 1-0 kHz tones. The overall ratio of highpitched to low-pitched tones was 1:9. The EEG signal was amplified by 10000 and passed through analogue filters which were set from 1 to 30 Hz (3 db down). The signal was then digitized at a rate of 1 kHz over a time epoch of 750 ms. A total of 75 ms of EEG was captured prior to delivery of the auditory stimulus to serve as a baseline for subsequent measurements. A total of 500 epochs was recorded and the data for high and low pitched tones averaged separately. An artefact reject circuit excluded all trials where the voltage exceeded + 45 /iV. Two recordings were made with each subject.
Recording of auditory event-related potentials Event-related potentials were recorded using a Nicolet CA1000 clinical signal averaging computer. Recordings were carried out exactly as in our previous studies (Blackwood et al. 1987, 1988; Kutcher et al. 1987; St. Clair et al. 1989; Muir et al. 1988) and the method is identical to that used by as number of other groups (Goodin
ERP rating Investigators who were blind to the subjects' diagnoses independently measured latencies and baseline-to-peak amplitudes of the individual components of the recorded waveforms. As in our previous studies these were designated as N l , P2, N2 and P3. Nl was the point of maximum negativity in the region 70—120 ms post-stimulus and P2 was the maximum posi-
Patients with affective disorders These subjects were selected on the basis of a clinical diagnosis of bipolar disorder or major depressive disorder. Eighty-eight met RDC criteria for bipolar disorder, of whom 75 were bipolar-I and 13 bipolar-II. Nine subjects met the criteria for bipolar-I disorder but were too ill (largely manic) to cooperate. Forty-eight patients met RDC criteria for major depressive disorder. In-patients with non-psychotic psychiatric disorders Thirty-two in-patients with non-psychotic disorders were used as a comparison group. None met the criteria for borderline personality disorder using the Diagnostic Interview for Borderlines (DIB; Gunderson et al. 1981). Details of the diagnoses of this mixed group of disorders are given in Table 2.
Long-latency auditory event-related potentials
Table 2. Diagnoses of non-psychotic psychiatric disordered group RDC Diagnosis
N
Minor depressive disorder 9 Intermittent depressive disorder 3 Anorexia nervosa 9 (3 also with minor depression) Bulimia nervosa Obsessive compulsive disorder Generalized anxiety disorder Panic disorder Passive-dependent personality disorder Unspecified personality disorder Substance abuse disorder
tivity between 140 and 230 ms in the averaged data of the low-pitched frequent stimuli. P3 was defined as the maximum positivity between 260 and 500 ms and the N2 the maximum negativity immediately preceding the P3 wave in the averaged responses to the high-pitched rare tones. The baseline amplitude was determined from the prestimulus record in each case, and amplitude of the individual components was measured with respect to this from the penrecorded tracings. Latencies were measured from the stimulus to the point of maximum amplitude of each wave component using the cursor provided by the computer on the visual display unit and by inspection of the pen-recorded trace. When the deflection did not form a sharp peak the latency was measured at the point of intersection of the tangents to the upgoing and downgoing slopes measured from the penrecorded trace by the method of Goodin et al. (1978). The mean of the two separate recordings was taken for the final results. Separate investigators, blind to the ERP measures made all the diagnoses. Clinical state at time of testing The degree of depression at the time of testing was assessed using Hamilton rating scores (Hamilton, 1960) in 61 of the subjects with bipolar disorder. Three subgroups were defined: 14 subjects were in a depressive phase of the illness with Hamilton scores of 17 or over (bipolar depressed); 20 subjects had Hamilton scores of 10 or less and did not fulfil the RDC criteria based on SADS-L evaluation for current manic or hypomanic episode (bipolar non-
871
Table 3. Characteristics of clinical subgroups of the affective disorders (For details of classification see text.) N
Mean age (s.D.)
Female:Male
Bipolar subgroups Bipolar depressed Bipolar non-depressed Bipolar manic
14 20 24
36-4(130) 33-3(110) 39-5(13-3)
7:7 9:11 7:17
Unipolar subgroups Unipolar depressed Unipolar non-depressed
22 11
38-9 (12-4) 34-0(12-2)
14:8 5:6
depressed); thirdly a group of 24 subjects fulfilled the RDC criteria for current manic or hypomanic episode (bipolar manic). A further 3 subjects had Hamilton scores between 10 and 17. Similarly 39 subjects were diagnosed as having major depressive disorder (RDC criteria based on SADS-L evaluation) and two subgroups of these were defined on the basis of Hamilton rating scores. A score of 17 or over was used to define a subgroup of 22 subjects who were clinically depressed at the time of testing (unipolar depressed) and this compared with 11 cases who scored 10 or below (unipolar nondepressed). There were a further 6 subjects who had Hamilton scores between 10 and 17. Demographic data on these subgroups are given in Table 3. Statistical analysis All results were analysed using the Statistical Package for the Social Sciences (SPSS-X, 1986). The ERP data followed approximately normal distributions and patient groups were compared using one-way analysis of variance (ANOVA) and the significance of the difference between the means was tested using Scheffe's procedure. Where appropriate analysis of covariance was also performed using age as a covariate. Within group comparisons were evaluated using Student's t tests.
RESULTS The groups differed significantly in age, (F(4,470) = 5-86; P< 0-0001). Subjects with major depressive (P < 0-01) and bipolar disorders (P
Controls No significant differences Schizophrenia > Controls Bipolar disorder > Controls Schizophrenia > Non-psychotic disorders Bipolar disorder > Non-psychotic disorders Schizophrenia > Major depressive disorder Schizophrenia > Controls Bipolar disorder > Controls Schizophrenia > Major depressive disorder Bipolar disorder > Major depressive disorder Schizophrenia > Non-psychotic disorders Bipolar disorder > Non-psychotic disorders
P
Controls Schizophrenia > Non-psychotic disorders Bipolars > Controls Schizophrenia > Controls Schizophrenia > Major depressives Schizophrenia > Non-psychotic disorders Bipolars > Controls Bipolars > Major depressives
P3
P < 0001
P< 0-001 P < 0001 P < 0001
P < 005
Table 10 b. ERP amplitudes (/JV + S.D. (TV)) of drug-free subject by group (Comparisons by Scheffe's procedure following ANOVA.)
Controls Schizophrenia Major depressive disorder Bipolar disorder Non-psycholic disorders Nl P2 N2 P3
Nl
P2
3-4±l-8(2U) 3 0 + 1-9 (46) 3-5 ±3-5 (35) 3-6 + 2-8(27) 3-6+1-7(20)
4-0+1-7(210) 3-2 + 1-4(46) 3-8+ 1-8 (35) 3-2+1-6(27) 3-4 ±1-7 (20)
N2
P3
3-6 + 2-4 (169) 2-6+1-5(38) 31 ±2-8 (31) 4-2 + 2-7(22) 31 + 1-9(15)
9-8 + 2-9(212) 6-0 + 2-8(44) 8-0 + 303(35) 6-5 + 2-7(26) 9-1+3-5(20)
No significant differences No significant differences No significant differences Schizophrenia < Controls Bipolars < Controls Schizophrenia < Non-psychotic disorders Major Depressives < Controls
1987) and indicate a refractoriness of P3 latency change to the patient's medication status or current mood state. P3 latency is delayed in both schizophrenics and bipolars in contrast to unipolar affective disorder. This provides psychophysiological support to the clinical distinction between bipolar and unipolar forms of depression, a concept which is further strengthened by the latest family studies on affective disorder (Andreasen et al. 1987). There is no evidence to suggest a physiological dichotomy between bipolar I and II subtypes. The methods used in our study have the advantage of being simple, easy to use and well tolerated by patients showing a wide variety of clinical states including those who were very disturbed. However, it does limit the psychological and physiological interpretation of the
P< P< P< P
