10.1192/bjp.164.4.487Access the most recent version at DOI: 1994, 164:487-493.BJP 

A K Malla and R M NormanProdromal symptoms in schizophrenia.

Referenceshttp://bjp.rcpsych.org/content/164/4/487#BIBLThis article cites 0 articles, 0 of which you can access for free at:

permissionsReprints/

permissions@rcpsych.ac.ukto To obtain reprints or permission to reproduce material from this paper, please write

to this article atYou can respond http://bjp.rcpsych.org/letters/submit/bjprcpsych;164/4/487

from Downloaded

The Royal College of PsychiatristsPublished by on July 6, 2014http://bjp.rcpsych.org/

http://bjp.rcpsych.org/site/subscriptions/ go to: The British Journal of PsychiatryTo subscribe to

British Journal of Psychiatry (1994), 164, 487—493

It is generally believed that the onset and recurrenceof psychotic symptoms in schizophrenia are oftenpreceded by other symptoms and types of behaviour,usually referred to as prodromal symptoms. In recentyears there has been more research designed to identifyany such prodromal symptoms and assess theirrelationship to the psychotic process in schizophrenia.There are at least two major reasons for this interest.First, if such prodromal symptoms could be identifiedand then reduced, this might prevent the occurrenceof full psychosis(Herz, 1985; Lukoff et a!, 1986;Jolley et a!, 1989). Second, the discovery of prodromal symptoms that are truly predictive ofimpending psychosis might provide additional insightinto the psychopathology and aetiology of schizophrenia (DeLisi et a!, 1986; Huber & Gross, 1989).

The empiricalinvestigationof prodromalsymptomsin schizophrenia is a relatively recent development.Before 1980, publications on the topic consistedmostly of case reports. Docherty et al (1978), in areview of a large number of case reports, suggestedthat there were five consistent stages in the development of psychoticsymptoms, the firstthreeconsistingof primarily non-psychotic symptoms and types ofbehaviour (typically reflecting emotional dysphoria)which are then followed by psychotic disorganisation.

More recent and systematic investigations ofprodromal symptoms in schizophrenia have usedvarious methods. Some have been based on reportsfrom patients and their families about symptoms and

types of behaviour which they perceive to haveoccurred before psychosis appeared (Herz & Melville,1980; Kumar eta!, 1989). These investigations haveshown that patients and family members oftenretrospectively report having observed a number ofnon-psychotic symptoms and/or certain alterationsin behaviour, which they believe to have precededany psychotic symptoms and behaviour. The mostfrequently cited possible prodromal symptomsconsist of mood changes such as tension, irritability,depression, anxiety, withdrawal, and vegetativechanges such as disturbed sleep and loss of appetite.

Information about the beliefs of psychiatrists andother mental health workers concerning prodromalsymptoms comes from studies evaluating low-doseor intermittent medications (Herz eta!, 1982, 1989,1991; Carpenter & Heinrichs, 1983; Heinrichs &Carpenter, 1985; Hirsch & Jolley, 1989; Hirsch eta!,1989;Jolley eta!, 1989, 1990). Such studieshave usedprotocols in which neuroleptic medications werealtered or reinstated when patients were judged tobe showing symptoms prodromal to psychosis.Typically, in these studies there was no directverification of the value of such symptoms as beingactual prodromes to psychosis; and so generally theirdata is best interpreted as reflecting beliefs aboutprodromal symptoms. The most common symptomstreated as prodromal were similar to those reportedby patients and their families and would fit thecategory of non-psychotic symptoms. However,

487

Prodromal Symptoms in SchizophreniaASHOK K. MALLA and ROSS M. G. NORMAN

Thispaperdescribesa prospectivestudyof the relationshipbetweennon-psychoticprodromalsymptomsand psychoticsymptomsin 55 schizophrenic(DSM—lll—R)out-patients.Once amonth, a numberof non-psychoticsymptomsgenerallyregardedas prodromalsymptomsinschizophreniawere assessed, as well as psychotic symptoms, with standardisedselfadministeredinstrumentsandratingscalesfor a minimumof 12 months(range12—29).Thedata were analysedfor each patientusinga longitudinalcorrelationaldesignwith a 1-monthlag between the prodromaland psychoticsymptomsover the total period.Resultsshowedthat in lessthan one-fifth of subjectsdid any of the prodromalsymptoms,individuallyor incombination,show a significantlypositivecorrelationwith the subsequentlevelof psychoticsymptoms.Suchrelationshipswere significantinanevensmallerproportionof subjectswhenthe confoundingeffect of concurrentpsychoticsymptomson prodromalsymptoms waspartialledout. Highlevelsof prodromalsymptomsappearedto haveadequatespecificitybutlow sensitivityintheirpowerto predicthighlevelsof subsequentpsychoticsymptoms.Therewere nodifferencesinage,gender,medicationlevels,andthe numberof previousadmissionsbetween the subjectswho did or did not show a relationshipbetween putative prodromalsymptomsand psychoticsymptoms.

488 MALLA & NORMAN

some investigators have also included symptomssuggestive of early exacerbation of psychosis, suchas hallucinations, inappropriate suspiciousness andthought disorder (Carpenter & Heinrichs, 1983;Heinrichs & Carpenter, 1985; Herz et al, 1989).

The value of classifying early signs of increasingpsychosis as prodromal is contentious (Herz &Simon, 1986). Generally in medical literature theterm ‘¿�prodromal'is applied in the field of infectiousdiseases to describe a set of symptoms that indicatethe initial stages of a disease: typically, generalisednon-specific symptoms such as malaise, fever, achesand pains, etc., which occur before the appearanceof the symptoms which are specific to a particulardisease (such as a unique pattern of skin eruptions).Such prodromes are indicative of the possible onsetof any one of a variety of infectious diseases. If weuse the term ‘¿�prodromalsymptoms' in an analogousfashion with regard to schizophrenia, then the natureof these symptoms must differ from the specific ordefining psychotic symptoms of schizophrenia. Itwould appear, therefore, to be more appropriate andless confusing to restrict the use of the term‘¿�prodromal'to non-psychotic symptoms.

Several longitudinal studies have measured changesin symptoms in patients suffering from schizophreniaand then retrospectively looked at the period beforea documented increase in psychotic symptoms forevidence of increases in possible prodromal symptoms. For instance, both Marder et a! (1984) andSubotnik & Nuechterlein (1988) have reportedevidence of a higher average level of such symptomsas depression, anxiety and/or somatic concern inperiods immediately before relapse than at othertimes. Such studies, however, fail to provide aprospective analysis of the relationship between nonpsychotic prodromal symptoms and psychosis, whichwould be crucial in assessing the predictive powerof such symptoms for future occurrence of psychosis.

The most direct evidence concerning the reliabilityof prodromal symptoms comes from studies whichassess their sensitivity and specificity as predictors ofpsychosis (e.g. Hirsch & Jolley, 1989; Birchwood et al,1989; Jolley et a!, 1990; Tarrier et al, 1991). Thesestudies have typically treated both prodromal and psychotic symptoms as dichotomous variables thaL areeither present or absent. The operational definition ofthe presence of each type of symptom varies betweenstudies, as do the number and types of possible prodromal symptoms examined and the length of time forwhich patients are followed. It is, therefore, notsurprising that there is considerable variation betweenthese studies in their estimates of sensitivity (from50% to 73%) and specificity (from 16°loto 100°lo)of prodromal symptoms as predictors of psychosis.

Several issues concerning prodromes of psychosisin schizophrenia have not been adequately addressedby previous research in this area. For instance, thepublished literature indicates that not all increasesin psychosis are preceded by increases in putativeprodromal symptoms, nor are all increases in suchsymptoms followed by increases in psychosis. It isunclear, however, whether this reflects differencesbetween patients in the extent to which changes innon-psychotic symptoms anticipate changes inpsychosis (some patients always showing prodromesand others never showing any) or whether for allpatients prodromes have only modest power aspredictors. In addition, although there is significantagreement regarding some of the broad categoriesof symptoms that are potential prodromes, there maywell be idiosyncrasies across patients in the particularsymptoms that are prodromal. The extent ofindividual differences in the nature and/or reliabilityof prodromal symptoms has not been adequatelyexamined in previous research.

A second question is whether prodromal symptomsreally precede increases in psychosis or are an integralpart of, or reaction to, a gradual psychotic process.There is evidence to suggest that dysphoria inparticular,a prominentputativeprodromalsymptom,may be an integral part of the psychotic processwhich occurs contemporaneously with psychosis ormay be secondary to psychosis (Bartels & Drake,1988;Sins et al, 1988; Barnes et al, 1989;Hirsch eta!,1989; Norman & Malla, l991a). This is particularlyimportant for our understanding of the relationshipbetween psychotic and non-psychotic symptomsin schizophrenia.

In this paper we describe a longitudinal study whichprospectively examines the relationship between anumber of non-psychotic symptoms and types ofbehaviour which are generally regarded as prodromalsymptoms in schizophrenia, and the subsequent levelof psychotic symptoms. As a better way of fmding outwhether such prodromal symptoms occur prior topsychosis or are actually a response to initial increasesin psychosis, both non-psychotic and psychoticsymptoms are assessed on a graded scale rather thanforced into a dichotomy of presence or absence(Falloon, 1984). For such assessments we used widelyaccepted and standardised measures of symptoms.In addition, the data are analysed in such a way thatindividual differences in the strength of associationbetween any prodromes and psychosis will be noticed.

Method

Subjects for this study were schizophrenic patients attendingout-patient treatment programmes in a teaching general

489PRODROMALSYMPTOMSIN SCHIZOPHRENIA

DisorganisationSAPS 23 - Aggressiveand agitated behaviourSAPS 26 —¿�DerailmentSAPS 27 - TangentialitySAPS 28 —¿�IncoherenceSAPS 29 —¿�IllogicalitySAPS 31 - Pressure of speechSAPS 32 —¿�Distractible speechSANS 6 —¿�Inappropriate affectSANS 10 - Poverty of content of speech

Reality distortionSAPS I - AuditoryhallucinationsSAPS 2 - Voices commentingSAPS 8 - Persecutory delusionsSAPS 11 - Grandiose delusionsSAPS 12 —¿�Religious delusionsSAPS 14 —¿�Delusions of referenceSAPS 15 —¿�Delusions of being controlledSAPS 16 —¿�Delusionsof mind readingSAPS 17 —¿�Thought broadcastingSAPS 18 - Thought insertion

Ratings on SAPS and SANS were conducted byinterviewerswho had receivedextensivetraining from theprincipal investigators (AM and RN). All ratings were madeon 6-point scales. In our research programme we havefound our inter-rater reliability, as estimated by theintraclass correlation coefficients (ICC), to be significantfor all items at a minimum of the 0.005 level. The actuallevel of the ICC varies across items on the two scalesbetween 0.61 and 0.98.

Data analysis

The data collected each month on both the non-psychoticand psychoticsymptomswereanalysedusinga longitudinalcorrelational design. These relationships were examinedwith a one-month lag between the prodromal and thepsychotic symptomsacross all time points for each subject.

Since subjects who show little or no variation in thelevel of psychotic symptoms over time cannot provideinformation regardingpossible prodromal symptoms, weselected for further study those patients who, duringfollow-up, demonstrated a range of scores on the realitydistortion (n = 24) and/or the disorganisation syndrome(n = 23) (both: n = 15) equal to, or greater than, I0°loofthe maximal potential range.

Scoreson eachof the non-psychoticsymptomspreviouslydescribedwerecorrelatedwithscoreson the realitydistortionand disorganisation syndromes, as observed one month later,for each subject across all monthly observations. Suchcorrelations indicate the extent to which the level of nonpsychoticsymptomspredictsthe levelat whichthe psychoticsymptoms (reality distortion and disorganisation) will beone month later. In the case of patients for whom therewas a significant serial correlation within a symptom overtime, the correction to the degrees of freedom suggestedby Holtzman(1963)wasadoptedfor testingthe significanceof correlationsof that symptomwithany other symptoms.

Partial correlations between prodromal symptoms andpsychoticsymptomsoccurring1 month laterwereconducted

hospital and two psychiatric hospitals. The subjects hadall been diagnosedas schizophrenic,confirmedthrough aStructuredClinicalInterviewfor DSM-III-R (Spitzereta!,1985) conducted by an experienced psychiatrist. They wereagedbetween17and 60,had at leastonepreviousadmissionto hospital for treatment of schizophrenia and had achieveda relatively stable clinical condition with neurolepticmedication for at least 3 months. Patients with a historyof organic brain disease, head injury, and/or significantdrug or alcohol dependence were excluded.

The subjectswereassessedeach month as partof a largerstudy of stress and symptoms in schizophrenia: 55completed monthly assessments for at least one year.Patients were enteredinto the study at different times, andthe total length of time they stayed in the study wasdetermined by funding considerations. The average numberof monthly assessments carried out was 19, with a rangefrom 12 to 29.

At the time of entry into the study, data werecollectedon each patient regarding age, gender, marital status,education, and employment status, as well as number ofpast admissions for psychiatric treatment, length of timesince discharge from hospital, and current neurolepticmedications. Information on treatment variables wascorroborated with clinicians and the patients' files.

In keepingwith the broadcategoriesreportedin previousliterature,the following symptoms and types of behaviourwere chosen as representing possible prodromal symptoms:depression, anxiety, somatic concern, feelings of beingstressed, low general functioning, and social withdrawal.Based on recent researchon the structureof symptoms inschizophrenia (Kulhara et a!, 1986; Liddle, 1987; Liddle& Barnes, 1990; Arndt et a!, 1991), as well as the resultsof our own factor analysis of symptoms from this patientpopulation (Malla et a!, 1993), the psychotic symptomswere grouped into two syndromes: reality distortion anddisorganisation. Previous work (includingour own), hasprovided evidence of a third syndrome, psychomotorpoverty, but we chose to exclude this from our analysis assome of the symptoms included in it could be regardedasprodromal and are not generally considered as part of thecore psychotic symptoms.

During each monthly assessment, symptoms wereassessed by a combination of self-report and rating scalescompleted by a trained interviewer. The Beck DepressionInventory (BDI; Beck, 1978)was used as a measure of thelevel of depression; anxiety was assessed with the SelfEvaluation Questionnaire (SEQ; Spielbergeret a!, 1968);and somatic concerns and level of general functioning wereassessed using the relevantsubscalesof the GeneralHealthQuestionnaire (GHQ.28; Goldberg, 1978). The PerceivedStress Scale (Cohen eta!, 1983; Norman & Malla, 1991b)was used to provide a global measure of subjective stress;and social withdrawalwas assessed through ratings on theasociality subscale of the Scale for Assessment of NegativeSymptoms (SANS; Andreasen, 1983).

The two psychotic syndromes were rated primarily onthe basis of relevant items of the Scale for the Assessmentof Positive Symptoms (SAPS; Andreasen, 1984) andspecific items from the SANS. The items from eachinstrument included to assess each syndrome were:

Non-psychotic symptomsPsychotic

Reality distortionsyndrome:%

DisorganisationDepression20.817.4Anxiety16.613Somatic

concern12.513Subjectivestress8.38.7Generalfunctioning4.24.3Social

withdrawal04.3

Non-psychotic symptomsPsychoticReality distortionsyndrome:

%DisorganisationDepression4.28.7Anxiety12.58.7Somatic

concern00Subjectivestress4.24.3Generalfunctioning4.24.3Social

withdrawal04.3

490 MALLA & NORMAN

after controlling for level of psychotic symptoms at theinitial assessment. This was done across all time points foreach individual subject. Further, in order to explore thepossibility that a combination of prodromal symptomsmay be more clearly predictive of subsequent psychoticsymptoms than are individual prodromal symptoms, weadded the standard score equivalences of each of theprodromes for each subject into a composite measure. Thiscomposite index was then correlated over time with the‘¿�realitydistortion' and ‘¿�disorganisation' syndrome scoresin the same manner as described above for individualprodromal symptoms. We also conducted additionalanalyses by defining substantial increases in prodromal and

psychotic symptoms as discrete events. This allowed anassessment of the sensitivity and specificity of the compositeindex of prodromal symptoms in the prediction ofsubsequent psychotic symptoms for each subject.

Results

Of the sample of 55 subjects, 72.7°7owere male and 27.3°lofemale, with ages ranging from 21 to 53 years (mean 37.8).The number of previous psychiatric admissions varied from1 to 9, with an average of 3. Their educational levels were:incomplete primary school, 3.6°7o;primary completed but

no secondary, 1.8070; incomplete secondary, 29.8%;secondary completed, but no post-secondary, 15.7%; postsecondary (college or university) but incomplete, 25.4%;post-secondary completed, 23.7°lo.Almost half (54.5%)were unemployed, 20°7oin casual and part-time employment,20% in full-time employment, 4 were students and one wasa housewife. Medication doses ranged from 20 to 2375 mgchlorpromazine equivalent per day, with an average of412 mg. This sample probably represents a subacute ormildly chronic patient population with an average numberof three psychiatric hospital admissions.

The percentage of correlations that were significant atthe 0.025 level (1-tailed) or better, for each of the prodromalsymptoms, are presented in Table 1. Substantially higherpercentages of significant correlations than would beexpected by chance were found when relating depression,anxiety, somatic concern and subjective stress to both realitydistortion and disorganisation. Of the 144 correlationscalculated regarding the reality distortion syndrome, 10.5%

Table 1Percentage of significant positive correlations between nonpsychotic symptoms and psychotic symptoms, with

1-month time lag

were significant at the 0.025 level and 8.3°loat the 0.005level. For the disorganisation syndrome, the correspondingfigures were 10. iWo and 5.8°lorespectively. Overall, theseresults certainly suggest a greater than chance level ofsignificant correlations between non-psychotic symptomsand subsequent levelof psychotic syndromes. However, thisrelationship only exists for a relatively small proportion ofcases. Correlations between composite indices of prodromalsymptoms and subsequent level of psychotic symptomsrevealed significant relationship in 23.1% of patients forreality distortion and l5.4°lofor disorganisation syndrome,respectively. Comparing these figures with those in Table 1shows that the composite index did not substantiallyimprove prediction beyond the level attained by the bestprodromal symptom - depression.

Of course the non-psychotic symptoms themselves maybe primarily a reflection of concurrent level of psychosisrather than an independent predictor of subsequentpsychosis. In order to assess this possibility we usedvariation in non-psychotic symptoms over time to predictpsychosis in the subsequent months, after partialling outthe variance in both that could be accounted for by the levelof psychosis at the time that the non-psychotic symptomswere assessed. A summary of these partial correlations ispresented in Table 2. A comparison of Tables 1 and 2suggests that a substantial proportion of the predictivepower of non-psychotic symptoms may, in fact, be areflection of their correlation with contemporaneousmeasures of psychosis. This is particularly noteworthy fordepression, somatic concerns, subjective stress, and, to alesser extent, anxiety. The effect of general functioning andsocial withdrawal on future psychotic syndromes, althoughinitially small, remains unchanged when the possibleinfluence of earlier psychotic symptoms is partialledout. In general these results suggest that much of thepredictive usefulness of the non-psychotic symptoms maybe because they themselves are related to early and subtlesigns of psychosis.

The relatively low proportion of significant correlationsbetween the putative prodromal and subsequent psychoticsymptoms may reflect either of two possibilities: (a) thatpsychotic symptoms are not frequently preceded byprodromal symptoms, but when the latter do occur, theyare generally followed by psychotic symptoms; or (b) thatwhen putative prodromal symptoms do occur they are oftennot followed by psychotic symptoms. In order to explore

Table 2Percentage of significant partial correlations between nonpsychotic symptoms controlling for previous level of

psychotic syndrome

491PRODROMAL SYMPTOMS IN SCHIZOPHRENIA

these possibilities both the prodromal and psychoticsymptoms need to be treated as discrete or categoricalvariables so that their presence or absence at a certain levelcan be defined. There are two additional reasons forcarryingout an analysisbasedon a categoricalmeasureofprodromal and psychotic symptoms. It might be argued thata strong relationship between prodromal and psychoticsymptoms is likely to be found only when there is asubstantial increase in these symptoms. Such an analysismay also be of practical value in that cliniciansgenerallytend to make decisions regarding treatment when theyobserve a significant or a substantial change in symptoms.

To explore these possibifitiesfurther, we analysed therelationshipbetweenhighlevelsof prodromaland psychoticsymptoms. We have avoided the use of hospitalisation asan index of a high levelof psychosisbecause experienceindicates that hospitalisation can occur for a myriad ofreasonsother than an increasein psychosis(Fallcon, 1984).Wechoseto definelevelsof both prodromaland psychoticsymptoms as being high when they were at least onestandard deviation above the relevant mean for eachsubject. We then calculated levels of sensitivity andspecificity of the effectiveness of prodromal symptoms inpredicting psychotic symptoms for each individual subject.The use of these indicesto assessthe predictivepower ofprodromal symptomshas been reported in severalrecentstudies (Hirsch & Jolley, 1989; Birchwood eta!, 1989; Jolleyeta!, 1990; Tarrier eta!, 1991). For this purpose we usedthe compositescore for prodromal symptoms(seeabove)as the predictorfor the realitydistortionand disorganisationsyndromes. The results indicate high levels of specificityfor prodromal symptomsas predictorsof realitydistortionand disorganisation for all subjects. The specificity wasalways 90% or higher for predicting reality distortion and83% or higherfor disorganisation.On the other hand, thesensitivity of the composite prodromal index as a predictor

of both syndromes was very low (less than 50% for 83°loof subjects).

A high specificity means that it is relatively rare thata substantial increase in prodromal symptoms is notfollowed by a substantial increase in psychotic symptoms(disorganisation or reality distortion). On the other hand,the comparatively low sensitivity indicates that manyincreases in psychotic symptoms are not preceded byincreasesin putative prodromal symptoms.

Discussion

The results of this provide some support for thenotion that a number of non-psychotic symptomsand behavioural changes are prodromal to the onsetof elevation of psychotic symptoms in the course ofschizophrenia. The predictive validity of theseprodromal symptoms appears, however, to belimited to a relatively small proportion of thepatients, particularly when the confounding effectof previous psychotic symptoms was partialledout. Depression appeared to be the best predictorof subsequent psychotic symptoms, although its

predictive validity was significantly less when theeffect of concomitant psychotic symptoms wasremoved. Assessing the relationship between thecomposite of all prodromal symptoms and subsequentpsychoticsymptomsdid not appear to improve uponthis relationship. However, when this relationshipwas further examined, treating prodromal andpsychotic symptoms as discrete events, the resultsshowed that whenever high levels of prodromalsymptoms occurred, they were likely to be followedby psychotic symptoms, but many increases inpsychotic symptoms were not preceded by increasesin prodromal symptoms. Therefore the non-psychoticsymptoms that were treated as putative prodromalsymptoms in this and in other previous studies doappear to have a specific value for subsequentpsychotic symptoms in those subjects who experiencesuch prodromal symptoms.

We have restricted our definition of prodromalsymptoms to those non-psychotic symptoms andbehaviours which have been reported in the literaturefor various samples of patients (Herz & Melville,1980; Marder et a!, 1984; Hirsch & Jolley, 1989;Jolley et a!, 1989; Tarrier et a!, 1991), and notincluded symptoms such as thought disturbance orhallucinations (Carpenter & Heinrichs, 1983;Subotnik& Nuerchterlein, 1988;Herz eta!, 1989). As indicatedearlier, we believe this distinction is conceptuallyfundamental to the understanding of the role ofprodromal symptoms in schizophrenia.

Most previous research on prodromal symptomsin schizophrenia has considered the onset ofpsychosis to be a discrete event of ‘¿�relapse'.As Herzet a! (1989) have suggested, the onset of psychosisshould not be considered an all-or-none phenomenon.It is much more likely to be a gradual processinvolving a progression from subtle disruptions inperception of reality and thought structure to moreblatant inability to identify reality or think coherently.Reports by Marder et a! (1984), Subotnik &Neuchterlein (1988), and Herz et a! (1991) suggestthat, if careful observations are made, gradualincreases in psychotic symptoms can be observed.In many cases, the supposedly prodromal signs mayreflect difficulty that the individual is having incoping with initial stages of psychosis, but thenon-psychotic symptoms may be more apparentthan the more subtle, and perhaps more private,psychotic symptoms.

Few studies have provided a true prospectiveanalysis of the relationship between prodromalsymptoms and psychosis. Subotnik & Nuechterlein(1988) provide initial prospective analysis of thepredictive power of only the early increases inthought disturbance, but not of other, non-psychotic

492 MALLA & NORMAN

prodromal symptoms. Our present paper reports onone of the few attempts to use prospective analysesto describe the longitudinal relationship betweennon-psychotic prodromal symptoms and psychoticsymptoms in schizophrenia. We have carefullyexamined this relationship in each subject at monthlyintervals for at least one year, and often longer.Correlations reported for each subject over time arelikely to provide more accurate information thancorrelations between subjects, as the variation invulnerability within subjects with time is likely to berelatively less than that between subjects. Furtheranalysis using higher levels of prodromal andpsychotic symptoms as discrete events has providedevidence in support of prodromal symptoms beingable to predict subsequent psychotic symptoms,although many psychotic episodes clearly occurwithout prior prodromal symptoms.

The relationships we have observed are unlikelyto have been confounded by changes in the patients'medication over time. Of the 55 patients, only four(707o) appear to have had changes in medication inresponse to increasing severity of symptoms.

The issue of whether prodromal symptoms trulyoccur prior to psychosis is probably of littleconsequence if they are used simply for the predictionof subsequent level of psychosis. It may, however,have implications for our understanding of schizophrenia. Docherty et a! (1978) proposed fiveidentifiable stages in the onset of psychosis inschizophrenia, extending from general feelings ofnervousness and being overwhelmed, to frankpsychosis. This is consistent with a number of modelswhich see psychosis as an end-point in a hierarchyof symptoms (e.g. Foulds & Bedford, 1975; Foulds,1976). Our findings suggest that dysphoria and otherputative early stages may not always occur beforeincreases in psychosis. Furthermore, even when suchnon-psychotic symptoms do appear to predictsubsequent psychosis, they may partially reflect earlyand subtle increases in psychotic symptoms.

We have contrasted our data for those patientswho did and did not show a time-lagged correlationbetween each non-psychotic symptom and psychosis.In these contrasts we have looked for evidence ofany significant differences in such characteristics asage, gender, number of previous admissions andchlorpromazine equivalence of their neurolepticmedication. We found no consistent evidence thatany of these variables are related to the likelihoodof prodromal symptoms. It may be that prodromalsymptoms are more consistent predictors of variationin level of psychosis than our data suggest. Patternsconsistent with a sequence of symptoms from nonpsychotic to psychotic may not be observed because

they occur more rapidly than can be detected by ourmonthly assessments. Another possibility is thatsome patients are displaying prodromal symptomsother than those we have assessed. It should be bornein mind that the potential prodromal symptomsassessed are consistent with suggestions from previousresearch, and that a one-month assessment schedulehas typically been used in previous research.

Conclusion

The study presented here differs from much previousresearch on prodromal symptoms of psychosis inschizophrenia in that symptoms were measured ona graded scale, the analysis was truly prospective,and we analysed the data so as to allow for theexamination of patterns of inter-relationship betweensymptoms for individual patients. Only a relativelysmall percentage of patients showed a significantcorrelation between non-psychotic symptoms and thelevel of psychotic symptoms one month later, and thispercentage was reduced even further when statisticallycontrolling for the level of psychosis at the time thenon-psychotic symptoms were assessed. However,additional analysis revealed that comparatively highlevels of prodromal symptoms do have the power topredict subsequent psychotic symptoms. Nevertheless,a large proportion of psychotic episodes appear tooccur without identifiable prior prodromal symptoms.

Acknowledgements

We are indebted to Dr Williamson, on the staff at LondonPsychiatric Hospital, Ontario, and the staff at the CommunityRehabilitation Program at Victoria Hospital, London, Ontario, fortheir assistance in this research project, and Ms Painter forsecretarial assistance.

References

ANDREASEN, N. C. (1983) The Scale for Assessment of Negative

Symptoms (SANS). Iowa City, Iowa: The University of Iowa.(1984) The Scalefor Assessmentof Positive Symptoms

(SAPS). Iowa City, Iowa: The University of Iowa.ARNDT,S., ALLIGER,R. J. & ANDREASEN,N. (1991) The distinction

of positive and negative symptoms: the failure of a twodimensional model. British Journal of Psychiatry, 158, 317-322.

BARNES,T. R., CURSON,D. A., LIDDLE, P. F., et al (1989) Thenature and prevalence of depression in chronic schizophrenicinpatients. British Journal of Psychiatry, 154, 486—491.

BARTELS,S. J. & DRAKE, R. E. (1988) Depressive symptoms inschizophrenia: comprehensive differential diagnosis. Comprehensive Psychiatry, 29, 467—483.

BECK, A. T. (1978) Beck Depression Inventory. New York:Harcourt Brace Jovanovich.

BIRCHWOOD, M., SMITH, J., MACMILLAN, F., et al (1989) Predictingrelapse in schizophrenia: the development and implementationof an early signs monitoring system using patients and families asobservers, a preliminary investigation. Psychological Medicine,19, 649—656.

493

CARPENTER,W. T. & HEI1@iaIcI4s,D. W. (1983) Early intervention,limited, targeted pharmacotherapy of schizophrenia. SchizophreniaBulletin,9, 533—542.

COHEN, S., KAMARCK, T. & MERMELSTEIN, R. (1983) A globalmeasure of perceived stress. Journal of Health and SocialBehaviour, 24, 385—3%.

DELIsI, L. E., CROW, T. J. & HIRscH, S. R. (1986) The ThirdBiannual Winter Workshop on Schizophrenia. Archives ofGeneralPsychiatry,43, 706—711.

DocHpslY, J. P., V@ KAMMEN,D. P., SIRI5,5.0. & MARDER,S. R.(1978) Stages of onset of schizophrenic psychosis. AmericanJournalof Psychiatry,135, 420—426.

FALLOON,I. R. H. (1984) Relapse: A reappraisal of assessment ofoutcome in schizophrenia. Schizophrenia Bulletin, 10,293-299.

Fouu)s, 0. & BEDFORD, A. (1975) Hierarchy of classes of personal

illness. Psychological Medicine, 5, 181—192.—¿� (1976) The Hierarchical Nature of Personal Illness. London:

AcademicPress.GOLDBERG, D. (1978) Manual of the General Health Questionnaire.

Windsor: NFER/Nelson.HEINRICHS, D. W. & CARPENTER, W. T. (1985) Prospective study

of prodromal symptoms in schizophrenic relapse. AmericanJournalof Psychiatry,142, 371—373.

HERZ, M. (1985) Prodromai symptoms and the prevention of

relapsein schizophrenia. Journal of ClinicalPsychiatry, 46, 22-25.—¿� & MELVILLE, C. (1980) Relapse in schizophrenia. American

Journalof Psychiatry,137, 801—805.—¿�, SzYM@st4sKI, H. V. & SIMoN, J. C. (1982) Intermittent

medication for stable schizophrenic outpatients: an alternativeof maintenance medication. American Journal of Psychiatry,139, 918—922.

—¿� & SIMON, J. C. (1986) Prodromal signs of schizophrenia

relapse (letter). American Journal of Psychiatry, 143, 115—116.—¿�, GLAzER, W., Mutz@, M., et al (1989) Treating prodromal

episodes to prevent relapse in schizophrenia. British Journal ofPsychiatry, 155 (suppl. 5), 123—127.

—¿�, GLAzSE, W. M., MOSTERT, M. A., et a! (1991) Intermittent

vs. maintenance medication in schizophrenia. Archives ofGeneralPsychiatry, 48, 333—339.

HIILScI1,S. R. & J0LLEY,A. 0. (1989) The dysphoric syndrome inschizophrenia and its implications for relapse. British Journalof Psychiatry,155 (suppi.5), 46-50.

—¿�, —¿�, BARNES, T. R. E., et a! (1989) Dysphoric and

depressive symptoms in chronic schizophrenia. SchizophreniaResearch,2, 259_2611

HOLTZMAN,W. H. (1963) Statistical models for the study of changein the single case. In Problems in Measuring Change (ed. C. W.Harris). Madison: University of Wisconsin Press.

HUBER, 0. & GRoss, 0. (1989) The concept of basic symptoms inschizophrenia and schizoaffective psychoses. Recenti Progressiin Medicini, 80, 646—652.

JOLLEY,A. 0., HIRscH, S. R., McRn@ac,A., et al (1989) Trial ofbrief intermittent prophylaxis for selected schizophrenic outpatients: clinical outcome at one year. British Medical Journal,298, 985—990.

—¿�, HIRscI, S. R., MORRISON, E., et a! (1990) Trial of brief

intermittentneurolepticprophylaxisforselectedschizophrenicoutpatients: clinical and social outcome at two years. BritishMedicalJournal,301, 837—842.

KULHARA, P., KOTA, S. K. & JOSEPH, S. (1986) Positive andnegative subtypes of schizophrenia: A study from India. ActaPsychiatrica Scandinavica,74, 353—359.

KuI@sAR,S., TiIAR@,R. & RAJKuPw, 5. (1989) Coping withsymptomsofrelapseinschizophrenia.EuropeanArchivesofPsychiatry and Neurological Science,239, 213-215.

LIDDLE, P. F. (1987) The symptoms of chronic schizophrenia: are-examination of the positive—negativedichotomy. BritishJournal of Psychiatry, 151, 145—151.

—¿�& BARNES, T. R. (1990) Syndromes of chronic schizo

phrenia.BritishJournalof Psychiatry,157,558-561.Luxorr, D., LIBERMAN,R. P. & NUECHTERLEIN,K. H. (1986)

Symptom monitoring in the rehabilitation of schizophrenicpatients. Schizophrenia Bulletin, 12, 578—602.

MALLA, A., NORMAN, R. M. 0., WILLIAMSON,P., et al (1993)Three syndrome concept of schizophrenia: a factor analyticstudy. Schizophrenia Research, 10, 143—150.

MARDER,S. R., V@ Pu1mN, T., MINTz, J., ci a! (1984) Maintenance therapy in schizophrenia: New findings. In Drug MaintenanceStrategiesinSchizophrenia(ed.J.M. Kane).Washington,DC:American Psychiatric Association.

NoRMAN,R. M. 0. & M@LA,A. K. (1991a) Dysphoric mood andsymptomatologyin schizophrenia.PsychologicalMedicine, 21,897—903.

—¿� & —¿� (1991b) Subjective stress in schizophrenic

patients. Social Psychiatry and Psychiatric Epidemiology, 26,212—216.

SIRI5, S. 0., ADAN, F., COHEN, M., ci a! (1988) Post psychotic

depression and negative symptoms: An investigation of syndromaloverlap. American Journal of Psychiatry, 145, 1532—1537.

SPIELBERGER, C. D., GORSUcH, R. L. & LusiseNE, R. (1968) SelfEvaluation Questionnaire. Palo Alto, CA: Consulting PsychologistsPress.

SPITZER, R. L. & WILLIAMS, J. B. W. (1985) Structured ClinicalInterviewfor DSM-III-R - Patient Version(SCID-P), 7/1/85.New York: New York State Psychiatric Institute, BiometricsResearch Department.

Suaomnc, K. L. & NUEcIrrERLEIN, K. H. (1988) Prodromal signsand symptomsof schizophrenicrelapse.Journal of AbnormalPsychology, 97, 405-412.

TARRIER,N., BARROwcLOUGH,C. & B@esAN, J. S. (1991) Prodromalsignsof relapsein schizophrenia.SocialPsychiatryandPsychiatric Epidemiology, 26, 157—161.

*Ashok K. Malla, MB, FRCP(C), Associate Professor, Department of Psychiatry, University of WesternOntario and Chief of Psychiatry, Victoria Hospital, 375 South Street, London, Ontario, N6A 4G5, Canada;Ross M. 0. Norman, PhD,Associate Professor, Department of Psychiatry, University of Western Ontario

*Correspondence

PRODROMAL SYMPTOMS IN SCHIZOPHRENIA

(First received March 1992, fina! revision March 1993, accepted June 1993)