Available online at www.sciencedirect.com

Comprehensive Psychiatry 53 (2012) 674–678www.elsevier.com/locate/comppsych

Predictive value of thought disorder in new-onset psychosisJames Wilcoxa,⁎, George Winokurb,1, Ming Tsuangc

aUniversity of Arizona, AZ, USAbUniversity of Iowa, IA, USA

cUniversity of California at San Diego, CA, USA

Abstract

Objective: This research addresses the relationship of formal thought disorder in the early stages of psychotic illness to the long-termoutcome of mental health many years later. The specific topic of concern was to evaluate the prognostic significance of thought disorder onthe severity of psychosis over time.Methods: Subjects with new-onset psychosis were evaluated on a variety of measures including education, physical health, Brief PsychiatricRating Scale scores. They were also given the Thought, Language, and Communication Scale to evaluate thought disorder. Subjects wereinterviewed again at 10 and 20 years to evaluate variations in outcome. Appropriate statistical methods were used to evaluate changes in thelevel of functioning over time.Results: Thought disorder was not unique to schizophrenia. Bipolar patients presented with significant positive thought disorder at the onsetof psychosis. Overtime positive thought disorder gradually improved in most patients. Negative thought disorder was more persistent,especially in subjects with schizophrenia. Initial psychosis with thought disorder characterized by poverty of content seemed to be associatedwith poor long-term outcome.Conclusion: Formal thought disorder can predict outcome in some cases of psychosis. Not all types of thought disorder have the sameprognostic implication. Positive forms of thought disorder (pressured speech, tangentiality) had no significant predictive value. Negativethought disorder (particularly poverty of speech and poverty of content) tend to predict a chronic, more unrelenting course of illness.Published by Elsevier Inc.

1. Introduction

The prediction of outcome for severe forms of mentalillness has been a topic of interest since the early days ofpsychiatry. The progression from early psychiatric symp-toms to chronic conditions remains a clinical mystery.Indeed, the causes of severe and persistent mental illness arestill poorly understood beyond the theoretical stages [1].Although risk factors such as genetic diathesis and substanceabuse give global information about disease, few actualsymptoms have been investigated rigorously over time [1-4].This is of concern in a discipline where diagnosticimpressions are largely made by assessing behavioralchanges. It is, therefore, of considerable importance to

Funded, in part, by NIMH grant MH60485 and the CHOPEEndowment.

⁎ Corresponding author. Tel.: +1 520 792 1450x6037.E-mail address: James.Wilcox2@VA.GOV (J. Wilcox).1 Deceased.

0010-440X/$ – see front matter. Published by Elsevier Inc.doi:10.1016/j.comppsych.2011.12.002

gather such information about the prognostic value ofsymptoms and conditions that present in the initial stages ofillness, whenever possible.

Early observations of prognostic signs for psychoticillness have been noted by the literature for nearly acentury [5,6]. In the past few years, there has beenrenewed interest in the study of psychosis among youngadults. Researchers search for variables that may predictthe persistence of disability among prodromal cases andvarious forms of schizotypy [6-9]. Psychiatric epidemiol-ogy has been very useful in the evaluation of earlysymptoms and prognosis [1-4,10]. This methodology isincreasingly relevant to assess risk factors in the prodroma[1,7,8]. Risk factors that predict conversion of first-episodepsychosis into schizophrenia include high levels of geneticloading, drug abuse, low levels of education, and lowsocioeconomic status [1,10-14]. A growing body ofevidence suggests a link between thought disorder,biological diatheses, and other known prognostic factorsoccurring during early psychosis [15-18].

675J. Wilcox et al. / Comprehensive Psychiatry 53 (2012) 674–678

Since the early work of Bleuler [19], Kraepelin [5,6], andMoukas et al [7], thought disorder has been considered amajor symptom of psychosis. Andreasen and Grove [20,21]have found that formal thought disorder has predictive valuein psychiatric diagnoses over several months. They havefound that negative thought disorder seems to predict theshort-term outcome of schizophrenia over several months.Patients with a poverty of speech and poverty of contentwere more likely to have a form of schizophrenia, whereaspositive forms of thought disorder were more commonamong the affective psychoses [5,6,13,14,21]. Severalinstruments to evaluate thought disorder presently exist[20-23]. These inventories appear valid and reliable and havebeen used in a variety of studies [20-23]. General syndromeinventory instruments have been suitable for replication andevaluation over time and have proven to be useful tools in thearea of psychiatric epidemiology [12-15,20-24]. Indeed, thefield of epidemiology has been very useful in areas ofvalidity and prognostic prediction [13-15].

This study was designed to look for prognostic variablesin early stages of psychosis and to follow subjects over timeto assess the relative value of demographic and symptomaticdata several years after the first onset of symptoms. We weremainly concerned with cases of new-onset psychosis (in thefirst 8 months) and to see how these patients progressed overthe following years. We concentrated our efforts on subjectswho were psychotic, regardless of their diagnosis.

2. Methods

This study was designed to be a prospective follow-up ofsymptoms in cases of new-onset psychosis. Our subjectswere collected from consecutive admissions to the Univer-sity of Iowa Psychiatric Hospital and clinic as well as localcommunity mental health centers. All subjects were olderthan 18 years at the time of their recruitment, and all met thecriteria for schizophrenia, schizoaffective disorder, or manicbipolar illness. Appropriate permission was obtained fromthe proper ethics committee or institutional review boardbefore the study, and all subjects gave informed consent toparticipate in a series of interviews. All were informed thatrecontact in 10 and 20 years was planned. No blood or tissuesamples were obtained from any subject. This yielded a totalof 188 cases: 68 with schizophrenia, 60 with schizoaffectivedisorder, and 60 with bipolar disorder in a manic state. Thesewere matched for sex and age with a community sample of200 controls who had no symptoms of psychiatric illness.

The psychiatric subjects were all considered to be “newonset cases” with less than 8 months of active symptoms. Ineach case, diagnoses were made by the primary investigator(J.W.) using the Research Diagnostic Criteria [24]. Thesecond author (G.W.) made diagnostic assessments on 5% ofthe subjects as a control of reliability. The interrateragreement was high, with a κ score of 0.92. Our studygroup has an excellent record of maintaining diagnostic rigor

in terms of both validity and reliability over time asevidenced by many studies involving psychiatric epidemi-ology (13, 14, and 15). We did initial recruitment for 3 yearsfrom 1993 to 1996. Our rate of participation was good, with86% of potential subjects actually agreeing to be in the study.

Each person in the study provided additional informationabout themselves. This included family history of psychiatricillness, occupational status, educational background, age ofonset of psychiatric symptoms, and marital status. Eachparticipant was evaluated using the Brief Psychiatric RatingScale (BPRS) [25] and the Scale for the Assessment ofThought, Language, and Communication (TLC) [20,21].The TLC was conducted by elucidation of responses duringan unstructured 45-minute interview conducted in themanner of Andreasen [20]. We conducted the TLC on thethird day of clinical care. The interviews were standardizedin a nonclinical format to prevent discussion of psychopa-thology. Each interview began by having the subject talkabout their self for as long as possible, usually 15 minutes.Ratings were done in a note form during the interviews andtransferred to score sheets immediately afterward. Astandardized, neutral list of topics was used to keep theinterviewer as blind as possible. Discussion of symptomswas avoided. It was our intention to assess each individual at10 and 20 years after the initial evaluation.

2.1. Statistical analysis

Statistical analysis used κ coefficient, χ2, and/or Fisherexact test, where appropriate. Student t test and Pearsoncorrelation were used in comparisons. Statistics were doneusing the SPSS program (SPSS, Chicago, IL) [26,27].Multiple regression analysis was conducted to evaluate therelative effects of each variable on the overall mean for TLCscores and BPRS scores at 10- and 20-year follow-up. Allstatistical tests were conducted 2 tailed using an α value of.05, unless otherwise noted.

3. Results

We found that at the onset of the study, subjects withschizophrenia and schizoaffective disorder had similarlyhigh rates of thought disorder by the TLC. As expected,controls expressed little or no thought disorder. The clinicalsubjects presented with varied degrees of thought disorder.Initial BPRS scores demonstrated high rates of impairmentin all 3 psychiatric groups at first evaluation (mean scoresof 86 for subjects with schizophrenia, 81 for subjects withschizoaffective disorder, and 78 for subjects with bipolardisorder). The initial TLC total mean scores were 33 forsubjects with schizophrenics, 32 for subjects with schi-zoaffective disorder, and 29 for subjects with bipolardisorder (Table 1).

At 10-year follow-up, we were able to contact 86% ofsubjects. We found that levels of thought disorder remainedsignificantly higher in patients with schizophrenia than in all

Table 1Occurrence of thought disorder between groups at initial evaluation

Groups df t P

Bipolar mania vs schizophrenia 1 0.91 .48Bipolar mania vs schizoaffective 1 0.98 .55Schizoaffective vs schizophrenia 1 0.38 .64

able 3ontrasts of thought disorder between groups at 10 years

ontrasted groups df t P

ipolar mania vs schizophrenia 1 3.34 .01ipolar mania vs schizoaffective 1 2.98 .01chizoaffective vs schizophrenia 1 0.39 .66

676 J. Wilcox et al. / Comprehensive Psychiatry 53 (2012) 674–678

other groups. Levels of thought disorder dropped consider-ably in patients with schizoaffective disorder and verysignificantly in subjects with bipolar disorder. Levels ofemployment remained low in subjects with schizophreniaand schizoaffective disorder but improved in subjects withbipolar disorder. Brief Psychiatric Rating Scale scoresremained high for schizophrenia and schizoaffective groupsat 10 years but declined significantly for those subjects withbipolar disorder (mean scores of 70 for subjects withschizophrenia, 68 for subjects with schizoaffective disorder,and 38 for subjects with bipolar disorder). The mean totalTLC scores were 23 for subjects with schizophrenia, 20 forsubjects with schizoaffective disorder, and 3 for subjectswith bipolar disorder (Tables 2 and 3).

At 20-year follow-up, we were able to contact 79% ofsubjects. We found that both the schizophrenia andschizoaffective groups continued to have prominent thoughtdisorder and elevated BPRS scores. These 2 groups also hadlow levels of employment and higher levels of psychiatrichospitalization. By contrast, the bipolar disorder group,however, no longer had significant amounts of thoughtdisorder and had significantly improved BPRS scores withsustained social improvement relative to the first 2 groups(lower rates of hospitalization and higher rates of employ-ment) at 20-year follow-up. Brief Psychiatric Rating Scalemean scores were 65 for subjects with schizophrenia, 63 forsubjects with schizoaffective disorder, and 23 for subjectswith bipolar disorder). The mean TLC scores were 19 forsubjects with schizophrenia, 18 for subjects with schizoaf-fective disorder, and 5 for subjects with bipolar disorder. Theinitial scores for poverty of thought content were correlatedwith long-term BPRS scores (r = 0.67, P b .01) and bothsocial and clinical outcomes (r = −0.56 and r = 0.64,respectively; P b .01) for employment and number ofhospitalizations. Particular negative types of thought disor-der, such as poverty of thought and poverty of content,strongly predicted poor outcome at both 10- and 20-yearfollow-up (F = 4.9 [P b .01] and F = 6.9 [P b .001],respectively) in a multiple regression analysis. The initialmean negative thought disorder scores (poverty of speech

Table 2Contrast of negative vs positive thought disorder at initial evaluation

Groups df t P

Bipolar manic vs schizophrenia 1 4.39 .01Bipolar mania vs schizoaffective 1 4.19 .01Schizoaffective vs schizophrenia 1 0.38 .68

TC

C

BBS

and poverty of content) were highly predictive of clinicaloutcome at 10- and 20-year follow-up (Tables 4 and 5).

4. Discussion

This project replicates a number of previous studies andbuilds upon the existing data. It also connects previousfindings in a new prospective. We feel comfortable reportingthat the symptoms normally assessed as “negative thoughtdisorder” are strong predictors of poor clinical outcome overthe lifespan of many people. A reasonable number of patientswere followed for considerable period and assessed by thesame clinician using validated scoring techniques. As notedin the past work, thought disorder is not limited toschizophrenia [18-21]. We did, however, find that 2 kindsof thought disorder were more common in schizophreniathan in schizoaffective disorder or bipolar disorder. Thesetypes were poverty of speech and poverty of thought content.We note that not all kinds of thought disorder have the samepredictive effect over time. Those best characterized as“positive,” including pressured speech, perseveration, stiltedspeech, and tangentiality, had little prognostic value. Thebest predictor of poor outcome was poverty of thoughtcontent (P b .001), and this was highly associated withschizophrenic diagnosis. Poverty of speech and illogicalityalso correlated with the diagnosis of schizophrenia and withpoor long-term outcome.

Thought disorder was not uniformly predictive ofoutcome. Positive symptoms seemed to fade away overtime and were less prominent at both 10- and 20-year follow-up in all groups. By contrast, the more negative symptomsseemed to persist in schizophrenia and were less relentingoverall. The concentration of negative formal thoughtdisorder in patients with poor long-term outcome isreminiscent of the descriptive phenomenology of earlyauthors [5,6,21]. The literature suggests that positivesymptoms were often associated with episodic syndromesand a generally better outcome, whereas negative symptomstended to be more chronic and debilitating [5,6,28-32].

able 4ontrasts of thought disorder (total TLC scores) between groups at 20 years

ontrasted groups df t P

ipolar mania vs schizophrenia 1 3.76 .01ipolar mania vs schizoaffective 1 3.02 .01chizoaffective vs schizophrenia 1 0.58 .72

TC

C

BBS

Table 5Comparison of TLC abnormalities among diagnostic groups at 20-year follow-up

Variable n at follow-up Normal (n = 158) Manic (n = 47) Schizoaffective (n = 47) Schizophrenia (n = 49)

n % n % n % n %

Poverty of speech 2 2 1 2 30 50 44 63Poverty of content 1 .5 0 0 28 48 35 49Pressure of speech 2 1 8 16 2 4 1 2Distractible speech 2 1 2 4 2 4 1 2Tangentiality 1 .5 2 4 2 4 0 0Derailment 9 5 9 18 2 4 2 3Incoherence 0 0 0 0 4 7 4 6Illogicality 0 0 0 0 2 4 2 4Clanging 0 0 0 0 0 0 0 0Neologisms 0 0 0 0 0 0 0 0Word approximations 4 2 4 7 3 5 5 7Circumstantiality 11 6 6 11 4 7 2 3Loss of goal 5 3 6 11 4 7 0 0Perseveration 6 3 10 18 4 7 0 0Echolalia 0 0 0 0 0 7 0 0Blocking 0 0 0 0 2 4 4 6Stilted speech 1 .5 2 5 0 0 0 0Self-reference 1 .5 2 5 0 0 2 3

677J. Wilcox et al. / Comprehensive Psychiatry 53 (2012) 674–678

Similarly, the poverty of speech and thought contentsuggests the symptoms of schizophrenia by Bleuler [19].The bipolar group, although having significant levels ofinitial thought disorder, tended to have forms of thinkingcharacterized by tangentiality, circumstantiality, and pres-sure of speech. These affective cases went on to have a moreepisodic course and better general outcomes on a varietyof measures.

A shortcoming of this study was the relatively smallnumber of subjects enrolled. This was offset by the long-term nature of the study and the use of reliable rating scalesfor repeated measures. We were fortunate to contact a highpercentage of our subjects at follow-up, with relatively littleattrition or loss of data. Another shortcoming of this studywas the fact that absolute blinding was not possible.Although every safeguard was used to ensure that theinterviewer was blind to the diagnosis, some symptoms areunavoidable to the trained eye; for example, manic patientstend to talk rapidly than normal. This de facto observationwas unavoidable in our arrangement. Some changes couldhave improved this project. At the onset of this study, weassumed that treatment availability would be equal for allsubjects because all were obtained from the same smallnumber of clinics. Post hoc analysis revealed that 81% of thesubjects obtained their ongoing care at their original points ofcontact, leading us to believe that the level of treatmentdecision making was fairly consistent. Variation in treat-ments can greatly affect long-term outcome in any study ofthis duration. Methods for incorporation of treatmentvariables into analysis should be done in further studies.

The importance of our findings is in the prediction of thecourse of psychotic illness by using data gathered early in theillness. A great interest exists in understanding how to matchprodromal cases of schizophrenia to treatment to reduce

morbidity. Many variables affect the course of an illness ofthis type. Genetic diathesis is certainly a primary focus forongoing study. The influence of drug abuse and socialvariables are also clear contributors to prodromal outcome insome cases. The study of specific presentations, that is,symptoms, is vital to the understanding and early interven-tion in cases of new-onset psychosis. This is likely to be truein the prodroma as well. Indeed, early intervention requiressome further degree of sophistication in the understanding ofthe evolution of symptoms over time. We urge continuedstudy of unique symptoms such as thought disorder to helpguide and understand the developing features of severepsychiatric disease states.

References

[1] Marshall M, Rathbone J. Early intervention for psychosis: update ofCochrane Database. Cochrane Database Syst Rev 2011(6):CD004718.

[2] Rosen C, Grossman LS, Bonner-Jackson A, Faull R. Diagnostic andprognostic significance of Schneiderian first-rank symptoms: a 20 yearlongitudinal study of schizophrenia and bipolar disorder. ComprPsychiatry 2011;52(2):126-31.

[3] Mancuso F, HoranW, Kern R, Green M. Social cognition in psychosis:multidimensional structure, clinical correlates, and relationship withfunctional outcome. Schizophr Res 2011;125(2-3):143-51.

[4] Larson M, Walker E, Compton M. Early signs, diagnosis andtherapeutics of the prodromal phase of schizophrenia and relatedpsychotic disorders. Exp Rev Neurother 2010;10(8):1347-59.

[5] Kraepelin E. In: & Barclay RM, editor. Dementia praecox andparaphrenia. Edinburgh: E. S. Livingstone; 1919.

[6] Kraepelin E. In: & Ross Diefendorf A, editor. Clinical psychiatry: atextbook for students and physicians, 7th edNew York: Macmillan;1915.

[7] Moukas G, Stathopoulou A, Gourzis P, Beratis IN, Beratis S.Relationship of “prodromal” symptoms with severity and type ofpsychopathology in the active phase of schizophrenia. ComprPsychiatry 2010;51(1):1-7.

678 J. Wilcox et al. / Comprehensive Psychiatry 53 (2012) 674–678

[8] Handest P, Parnas J. Clinical characteristics of first-admitted patientswith ICD-10 schizotypal disorder. Br J Psychiatry 2005;48:49-54supplement.

[9] Woods SW, Addington J, Cadenhead KS, Cannon T, Cornblatt B,Heinssen R, et al. Schizophr Bull 2009;35(5):894-908.

[10] Keshavan M, DeLisi L, Seidman L. Early and broadly definedpsychosis risk mental states. Schizophr Res 2011;1126(1-3):1-10.

[11] Fumero A, Santamaria C, Navarrete G. Predisposition to alcohol anddrug consumption in schizophrenia-vulnerable people. Rev Neurol2009;49(1):8-12.

[12] Veen ND, Selten JP, Van der Tweet I, Feller WG, Hoek HW, Kahn RS.Cannabis use and age of onset of schizophrenia. Am J Psychiatry2004;161(3):501-6.

[13] Winokur G, Morrison J, Clancy J, Crowe R. The Iowa 500: familialand clinical findings favor two kinds of depressive illness. ComprPsychiatry 1973;14(2):99-106.

[14] Morrison J, Clancy J, Crowe R, Winokur G. The Iowa 500; I.diagnostic validity in mania, depression, and schizophrenia. Arch GenPsychiatry 1972;27(4):457-61.

[15] Morrison J, Winokur G, Crowe R, Clancy J. The Iowa 500; the firstfollow-up. Arch Gen Psychiatry 1973;29(5):678-82.

[16] Ettinger U, Kumari V, Crawford T, Flak V, Sharma T, Davis R, et al.Saccadic eye movements, schizotypy and the role of neuroticism. BiolPsychol 2005;68(1):61-78.

[17] Uhlhaas P, Silverstien S, Phillips W, Lovell P. Evidence for impairedvisual cortex processing in schizotypy with thought disorder.Schizophr Res 2004;68(2-3):249-60.

[18] Ruhrman S, Schultze-Lutter F, Salokangas RK, Heinimaa M, LinszenD, Dingemans P. Prediction of psychosis in adolescents and youngadults at high risk: results from the prospective European prediction ofpsychosis study. Arch Gen Psychiatry 2010;67(3):241-51.

[19] Bleuler E. In: & Zinkin J, editor. Dementia praecox, or the group ofschizophrenias. (1911). New York: International Universities Press;1950.

[20] Andreasen N. The Scale for the Assessment of Thought, Language andCommunication (TLC). Iowa City: The University of Iowa; 1978.

[21] Andreasen N, Grove W. Thought, language, and communication inschizophrenia: diagnosis and prognosis. Schizophr Bull 1986;12(3):348-59.

[22] Liddle PF, Ngan ET, Caissie SL, Anderson CM, Bates AT, QuestedDJ, et al. Thought and Language Index: an instrument for assessingthought and language in schizophrenia. Br J Psychiatry2002;181:326-30.

[23] Barrera A, McKenna PJ, Berriros GE. Two new scales of formalthought disorder in schizophrenia. Psychiatry Res 2008;157(1-3):225-34.

[24] Spitzer R, Endicott J, Robins E. Research diagnostic criteria: rationaleand reliability. Arch Gen Psychiatry 1978;35:778-82.

[25] Overall JE, Gormann DR. The Brief Psychiatric Rating Scale. PsycholRep 1962;10(2):799-812.

[26] Abraham B, Ledolter J. Statistical methods for forecasting. New York:John Wiley & Sons; 2005.

[27] Leversque R. SPSS programming and data management: a guide ofSPSS and SAS users. 4th ed. Chicago: SPSS Inc; 2007.

[28] Cannon TD, Cornblatt B, McGorry P. The empirical status of the ultrahigh-risk (prodromal) research paradigm. Schizophrenia Bulletin2007;33(3):661-4.

[29] Triptish B, Satabdi C, Pramod T, Amina N, Sati M, Vishwajit L. Isfamiliarity associated with downward occupational drift in schizo-phrenia? Psychiatry Invest 2008;5(3):168-74.

[30] SameleC,VanOsJ,McKenzieK,WrightA,GilvarryC,MankeyC.Doessocioeconomic status predict the course and outcome in patients withpsychosis? SocPsychiatryPsychiatr Epidemiol 2001;36(12):573-81.

[31] Waring EM, Neufeld RW, Schaefer B. The thought disorderquestionnaire. Can J Psychiatry 2003;48(1):45-51.

[32] Mercado CL, Johannesen JK, Bell MD. Thought disorder severity incompromised, deteriorated, and preserved intellectual course ofschizophrenia. J Nerv Ment Dis 2011;199(2):111-6.