Background

There are three strands to the literature on pre-illness features associated with schizophrenia. First, it has been well established that in childhood and the years before evidence of schizophrenia emerges, individuals differ from normal controls on a range of measures, including psychological test performance and patterns of behaviour (Reference Baum and WalkerBaum & Walker, 1995; Reference JonesJones, 1997). Such findings frequently are regarded as evidence that schizophrenia is a disorder of neuro-development that arises early but in some way is compensated until the typical age of clinical onset in young adulthood (Reference Murray and LewisMurray & Lewis, 1987; Reference WeinbergerWeinberger, 1987). Second, using retrospective patient and third-party accounts, it has been shown that psychopathology of various kinds can precede the emergence of diagnosable illness by months or years (Reference ChapmanChapman, 1966; Reference Hafner, Maurer, Loffler, Hafner and GattazHafner et al, 1995). Some authors have concluded that the psychotic shift is driven by affective change, either anxiety or depression (Reference Birchwood, Iqbal, Wykes, Tarrier and LewisBirchwood & Iqbal, 1998; Reference Garety, Kuipers and FowlerGarety et al, 2001), whereas others have concluded that affective change itself is consequent upon disturbances in either cognition or perception (Reference ChapmanChapman, 1966), something that would be difficult to determine with even detailed retrospective techniques. Third, greater emphasis has been given recently to minor psychotic or psychotic-like phenomena such as referential ideas, perceptual disturbances and magical thinking, in an attempt to determine the point at which early treatment, especially with antipsychotic drugs, would be appropriate (Reference McGlashan and JohannessenMcGlashan & Johannessen, 1996; Reference McGorryMcGorry, 1998; Reference McGorry, Yung and PhillipsMcGorry et al, 2002; Reference Woods, Breier and ZipurskyWoods et al, 2003).

Although the onset of schizophrenia can be very acute, it is often more gradual and the point at which symptomatology could be regarded as predictive or prodromal, rather than representative of the early features of illness itself, is often far from clear (Reference Beiser, Erickson and FlemingBeiser et al, 1993). Prospective population-based studies utilising the controlled and masked assessment of prepsychotic states would be impractical, but the possibility of such assessment has arisen within the context of the Edinburgh High Risk Study of schizophrenia.

The purpose of the present study is to relate initial symptomatic assessments of the high-risk participants and controls to those characteristic of patients already ill with a first episode of schizophrenia, with a view to considering whether non-psychotic symptoms are secondary to developing psychosis and to define the characteristics of the pre-illness state in high-risk individuals who eventually progress to an acute schizophrenic illness.

Analysis

The PSE data on participants entering the Edinburgh High Risk Study were available at entry on 175 individuals, divided into 127 high-risk participants who remained well through follow-up, 21 who developed a first formal schizophrenic illness (i.e. high-risk ill participants) and 27 normal controls. One-way analyses of variance (ANOVAs) were run, comparing these three groups and the Northwick Park first-episode patients on all the log-transformed symptom scales. Follow-up planned comparisons, not assuming equal variances, compared the 21 high-risk ill participants with the other Edinburgh High Risk Study groups and also to the Northwick Park patients. Non-parametric Kruskal–Wallis tests also were run and in all cases they confirmed the overall parametric findings. Possible gender effects were examined, using χ²-tests to assess group composition and two-way ANOVAs to search for interaction effects. A separate comparison was made of the Northwick Park patients with high-risk participants who fell ill and were assessed at the time of illness onset. A second analysis examined changes in symptoms between the first and last assessments for the Edinburgh High Risk Study participants. This was attempted using repeated-measure ANOVAs, with group (high-risk ill, high-risk well and control) as a between-participants variable.

Methodology

The methodology adopted in this study is unusual and the reasons why it was adopted require explanation. At the planning stage, the primary objective of interest lay in establishing ‘caseness’/‘non-caseness’, an aim to which the PSE as traditionally used is eminently suited. It subsequently became clear that symptomatology at entry was more prominent and less group specific than the literature or our own predictions had led us to believe and that it was desirable to make more detailed measurements. In the absence of other measures sensitive to symptom change it was decided to adapt the PSE for this purpose.

The PSE is an extensive instrument comprising 140 elicited and observed mental state phenomena, most of which are recorded as continuous variables along a range of ‘absent’/‘mild’/‘severe’, and it proved a relatively straightforward matter to construct the 12 scales with high α coefficients (see below and Appendix). In many ways they are similar to other measures of mental state and would be expected to be reliable, valid, sensitive to change and inclusive. The PANSS now has been incorporated into the study and the new PSE scales will be compared with these ratings at completion.

The 12 scales were derived from one of the largest samples of patients with first-episode schizophrenia to which the PSE was applied on admission (i.e. the North-wick Park sample). These patients were early in the course of their florid illness and for the most part only briefly exposed or not exposed at all to psychotropic medications. Because the PSE applies to the previous 4 weeks, it is unlikely that medication or other factors relating to admission significantly altered the ratings in this sample, which may be taken as generally representative of the illness in its acute state. This sample was preferable to the first-episode patients participating in the Edinburgh High Risk Study, in view of its much larger size and the fact that the Edinburgh patients mainly had been exposed to significant antipsychotic drug treatment, sometimes for months, by the time their PSEs were conducted.

The concept of schizophrenia has not changed radically in the 20 years separating the two studies, and the PSE remains a valid instrument. However, different raters were involved in the two studies, raising the question of reliability. Both E.C.J. and D.G.C.O. conducted PSEs in both studies and did the great majority in the second phase of the Edinburgh High Risk Study. In addition, for many years they ran a PSE training course, the first such course to be approved outside the Institute of Psychiatry. Research colleagues all received their clinical training in the same institutions as those in which the principal authors worked – a factor known to improve reliability (Reference KendellKendell, 1975) – and in addition underwent similar PSE training.

Thus, notwithstanding its limitations, we believe that the material presented here is reliable and valid.

Pre-illness symptomatology

This study shows that high levels of non-specific, affective symptoms are evident in patients with first-episode schizophrenia substantially before the onset of psychosis and that these separate those high-risk individuals destined to develop schizophrenia from the other high-risk individuals who remain well. These non-specific abnormalities remain essentially stable over time (and in most instances by ‘time’ we mean more than 2 years), both in those who progress to illness and in those remaining well, despite a non-significant tendency for ‘situational anxiety’ to diminish and ‘depression’ to exacerbate in those who progress to illness. These results suggest that such non-specific affective symptomatology is not merely secondary to emerging psychosis but is more fundamental to the illness process that it antedates. Furthermore, entry scores comprising ‘situational anxiety’ were significantly higher in those destined to progress to psychosis than in the sample who had been diagnosed with a first episode of schizophrenia (the North-wick Park sample), supporting the view that anxiety-type phenomena may partially remit as psychotic features escalate. Although this study cannot address the question of whether a greater risk accrues from anxiety (Reference Garety, Kuipers and FowlerGarety et al, 2001; Reference Turnbull and BebbingtonTurnbull & Bebbington, 2001) or depression (Reference Birchwood, Iqbal, Wykes, Tarrier and LewisBirchwood & Iqbal, 1998), a key role for anxiety is in keeping with other results from the Edinburgh High Risk Study, in which the best predictors of illness from mothers’ accounts recorded on the Childhood Behaviour Checklist (Reference Achenbach, Howell and QuayAchenbach et al, 1991) were withdrawn and deviant behaviour, which includes anxiety and depression (Reference Miller, Byrne and HodgesMiller et al, 2002). Using different measures, participants in the Israeli high-risk study who subsequently progressed to a schizophrenic-spectrum diagnosis were found to have had higher pre-illness levels of anxiety assessed at age 16 years (Reference Kugelmass, Faber and IngrahamKugelmass et al, 1995). Thus, anxiety phenomena may be an inherent part of the pathophysiological process mediating the schizophrenic syndrome.

The more specific symptomatology associated with pre-illness states relates to perceptual abnormalities and comprises both distortions and deceptions. Although ‘changed perception’ and ‘hallucinations’ were found in normal controls and high-risk participants who remained well, hallucinations in particular were infrequent but were more evident at entry in those who progressed to illness. Abnormalities of thought form and content did not differ significantly between the groups at entry. Group scores were very low and these features were found only in those destined to become ill. Together they comprised the major changes associated with formal illness development.

Premorbid or prodromal?

In the high-risk population, the mean time to illness onset was 929 days (s.e.=138). Because it has been reported that prodromal symptomatology can be present for many years prior to formal diagnosis (Reference Hafner, Loffler and MaurerHafner et al, 1999), this raises the question of what type of phenomenology the high-risk participants who fell ill were exhibiting at entry – premorbid or prodromal.

Interest in the pre-diagnostic phenomena associated with schizophrenia is long-standing but has increased markedly in recent years. Based on the wish to introduce earlier treatment that may have a favourable impact on outcome, attempts have increased to delineate the prodromal phase of illness from both its premorbid characteristics and the features of the florid first psychotic episode. There is a wealth of evidence that schizophrenia is associated with a wide range of premorbid deviations evident in a series of behavioural, neuro-psychological and even brain structural domains – observations confirmed in the Edinburgh High Risk Study sample (Reference Lawrie, Whalley and KestelmanLawrie et al, 1999; Reference Cosway, Byrne and ClaffertyCosway et al, 2000). These are essentially stable characteristics that do not necessarily result in disadvantage and are certainly not viewed as ‘clinical’ phenomena. What is less clear is how the illness prodrome (representing the first shift from the premorbid state towards illness) should be conceived: what features it comprises, how it evolves and where its ‘break’ points lie between normality and illness. This is an increasingly important question because present conceptualisations of prodrome, essentially derived from retrospective methodologies, have led to the advocacy of early treatment interventions with antipsychotic drugs, often in very young individuals. A significant impact on progression to psychosis has yet to be reported extensively (e.g. Reference McGorry, Yung and PhillipsMcGorry et al, 2002) and it remains unclear that very early interventions do result in better long-term outcomes.

A major problem is the definition of ‘prodrome’ in the context of schizophrenia, which of necessity is a retrospective concept (Reference Yung and McGorryYung & McGorry, 1996; Reference Cornblatt, Lencz and KaneCornblatt et al, 2001), whose constructs have arisen largely on the basis of interviews with patients already diagnosed and, in more recent work, with their families, supplemented with reference to medical records. Although this methodology may produce systems of assessment that are reliable (Reference Hafner, Loffler and MaurerHafner et al, 1999), the sources of bias continue to challenge their validity. Using these methods, current views of the contents of the prodrome do include a prominent place for non-specific symptomatology, including affective features, as reported here.

Chapman, on the basis of patient interviews conducted within 3 years of a first episode, found ‘intense anxiety’ to be ‘almost invariable’ and also that perceptual disorders were common, something he placed in a key role in his theory on the origins of florid symptomatology (Reference ChapmanChapman, 1966). This also would be compatible with our finding that, on entry, those who eventually became unwell demonstrated higher levels of perceptual abnormalities than the other two Edinburgh High Risk Study groups.

Thus, on this evidence, our high-risk sample destined for illness may indeed have been ‘prodromal’ at entry. However, no matter how the schizophrenic prodrome is conceptualised, some element of change from a previous state (essentially behavioural, but also subjective) is inherent to the concept (Reference Keith and MatthewsKeith & Matthews, 1991; Reference Hafner, Riecher-Rossler and HambrechtHafner et al, 1992; Reference Loebel, Lieberman and AlvirLoebel et al, 1992; Reference Beiser, Erickson and FlemingBeiser et al, 1993; Reference Yung and McGorryYung & McGorry, 1996; Reference Cornblatt, Lencz and KaneCornblatt et al, 2001). This key criterion did not apply to the participants of the Edinburgh High Risk Study, who were selected specifically on the basis of being well at entry, in both their own and their families’ eyes. Because these individuals came from families in which at least two members were already affected with schizophrenia, we take this information, especially that from family sources, as sound.

This might suggest that change need not be an inherent part of the schizophrenic prodrome, which if true would make the concept more arbitrary and difficult to pin down clinically than is currently believed. An alternative proposal might be that high levels of affectivity and perceptual aberration can, in a stable behavioural context, represent part of the premorbid state, perhaps the result of a gradual process of adaptation to underlying cognitive deficits.

Derivation of PSE scales

The PSE data from the Northwick Park Study of first episodes of schizophrenia were reduced to those who received CATEGO diagnoses of S+/S?, P+/P? or O+/O?, corresponding to schizophrenic psychoses, paranoid psychoses or ‘other’ psychoses (Reference Wing, Cooper and SartoriusWing et al, 1974). The protocols from the resulting 229 patients were used in the derivation of 12 scales.

The PSE is set out in 20 sections: 17 sections comprising symptomatology elicited by formal questioning; and 3 sections for recording mental state features observed during interview. Within these sections most of the items are rated on one of three anchor points, from absent to severe.

The scales were constructed in four stages.

Stage 4

Cronbach's α coefficients for each of the 12 scales were reassessed in a mixed sample of patients comprising 26 with first-episode schizophrenia in hospitals local to Edinburgh in 1994, 19 high-risk patients who fell ill (assessed at the time of their illness) and 98 patients from Northwick Park with psychoses not classified as schizophrenia.

The details of the final scales derived are set out in Table A1.

Table A1 Composition of the symptom scales

Scale	Items included	Derivation α (n=229)1	Confirmatory α (n=143)2
Situational anxiety	14. Panic attacks
	15. Situational autonomic anxiety
	16. Autonomic anxiety on meeting people
	17. Specific phobias	0.73	0.78
	18. Avoidance of anxiety-provoking situations
Nervous tension	4. Worrying
	5. Tension pains
	6. Tiredness
	7. Muscular tension
	8. Restlessness
	10. Feeling of nervous tension	0.75	0.76
	11. Free-floating autonomic anxiety
	12. Anxious foreboding with autonomic accompaniments
	120. Observed anxiety
Depression	19. Inefficient thinking
	20. Poor concentration
	21. Neglect due to brooding
	22. Loss of interest
	23. Depressed mood
	24. Hopelessness
	25. Suicidal plans or acts
	27. Morning depression
	29. Self-depreciation
	32. Guilty ideas of reference
	33. Pathological guilt
	34. Loss of weight due to poor appetite	0.80	0.85
	35. Delayed sleep
	36. Subjective retardation
	37. Early waking
	121. Observed depression
Mania	41. Expansive mood
	42. Ideomotor pressure	0.88	0.91
	43. Grandiose ideas
Over-activity	111. Agitation
	112. Gross excitement
	113. Irreverent behaviour
	122. Histrionic
	123. Hypomanic affect	0.66	0.80
	127. Lability of mood
	131. Pressure of speech
Disorganisation	114. Distractibility
	115. Embarrassing behaviour
	116. Mannerisms and posturing
	126. Perplexity
	129. Incongruity of affect	0.66	0.65
	132. Non-social speech
	135. Neologisms
	136. Incoherence of speech
Changed perception	47. Derealisation
	48. Depersonalisation
	50. Heightened perception
	51. Dulled perception
	52. Changed perception	0.71	0.41
	53. Déjà vu
Hallucinations	60. Non-verbal auditory hallucinations
	62. Voices discussing the patient
	63. Voices speaking to the patient
	64. Dissociative hallucinations
	65. Mind or ears	0.69	0.71
	66. Visual hallucinations
	68. Olfactory hallucinations
	70. Other hallucinations
Disorder of possession of thought	55. Thought insertion
	56. Thought broadcast
	57. Thought echo
	58. Thought block	0.62	0.53
	59. Thoughts being read
Delusional construction	72. Delusions of reference
	73. Delusional misinterpretation
	74. Delusions of persecution
	80. Physical forces
	93. Systematisation of delusions	0.79	0.85
	94. Evasiveness
	95. Preoccupation with delusions and hallucinations
	96. Acting out delusions
Outside control	71. Delusions of control
	75. Delusions of assistance
	76. Delusions of grandiose abilities
	78. Religious delusions
	79. Paranormal phenomena	0.71	0.66
	81. Control by alien forces
	82. Primary delusions
	92. Delusions of catastrophe
Negative symptoms	108. Self-neglect
	110. Slowness
	119. Catatonic movements
	128. Blunted affect	0.65	0.66
	130. Slow speech
	133. Muteness
	134. Restricted quantity of speech
Items not included	1, 2, 3, 9, 13, 26, 28, 30, 31, 38, 39, 40, 44-46, 49, 54, 61, 67, 69, 77, 83-91, 97-107, 109, 117, 118, 124, 125, 137-140

CLINICAL IMPLICATIONS

1. ▪ In the genetically predisposed, those who develop schizophrenia have prominent affective and perceptual symptomatology before evidence of the change in behaviour or subjective mental state conventionally used to delineate onset of prodromal illness.

2. ▪ In symptomatic terms, the prodromal phase of illness may be difficult to delineate from the stable premorbid state using retrospective methodologies.

3. ▪ The prominence of affective phenomenology raises the possibility of early interventions other than antipsychotic drugs.

LIMITATIONS

1. ▪ The method of data analysis in this study is novel and requires validation.

2. ▪ The use of data from samples acquired many years apart raises questions of reliability.

3. ▪ Although data on the high-risk participants were acquired prospectively, the number progressing to a first schizophrenic illness was relatively small.