Satria Adi

Schizophrenia is a diverse disorder. One facet of its diversity is theage at onset. The incidence of schizophrenia peaks between 10 and25yearsformenandbetween25and35for women.1Anotherpeak, particularlyamongwomen, occursinmid-life: about23%of peoplewithschizophreniaexperiencetheirfirst episodeaftertheageof40.2Inasmall groupofpeople, schizophreniahasitsonset after age 60, which has been defined as very-late-onsetschizophrenia-likepsychosis.3Finally, onsetofschizophreniacanoccur inchildhoodor adolescence, typicallyafter theageof 5.Its prevalence is about 1 per 10 000 inchildren, and12 per1000 in adolescents. As in early adulthood, higher rates arereportedinyoungmalesthaninyoungfemales.4DeLisi hasconceptualised age at onsetas a surrogate measureof severity of the disease process.5An earlier age at onset has beenassociated with more severe clinical and behavioural symptoms,6,7moresocial disability,8narrower posterior brainsegments9andlarger ventricles.10Cognitive deficits inschizophrenia are corefeatures of the illness.11,12They are the strongest predictors offunction and are considered potential targets for treatment.13They have beenwell characterised, for example by Heaton etal,14Hoff et al,15andWhiteet al,16andarebelievedtobethemanifestation of the disease process affecting different brainsystems. Thus, characterising the relationship between age at onsetandcognitivedeficitswillilluminatefromacognitiveperspectivethe unravelling of the disease process at different levels of severity.Although some studies have reported an association of earlier ageatonsetandmoreseverecognitivedeficits,6,1719othersfailedtofinddifferencesinthecognitiveprofilesofindividualswithearlyv. late-onset schizophrenia.14,20,21Further, the nature of thecognitive deficits that maybe associatedwithage at onset hasvaried in different studies. Considering that several of thesestudies have small sample sizes, and to assess whether suchassociationsexistacrossthelife-span, wepresentameta-analysisof the literature, estimating and comparing the severity ofcognitive deficits in patients with adult-onset at their first episodeof schizophrenia, youth-onset (in childhood or adolescence)schizophrenia and late-onset schizophrenia. We chose to compareindividualswithyouth-onsetorlate-onsetschizophreniaonlytopeople with adult-onset schizophrenia at their first episodebecause individuals with youth-onset and late-onset schizophreniaare typically studied close to the onset of the illness. By contrast, inindividualswithadult-onsetschizophreniawhohavebeenill formany years, one would expect chronicity and treatment toinfluencecognitivedeficits, thus, confoundingthehypothesisedspecificrelationships toageat onset. Inthis report weusetheword schizophrenia to refer to schizophrenia and relateddisorders: schizoaffective, schizophreniform, delusional disorderorvery-late-onsetschizophrenia-likepsychosis.MethodLiteraturesearchA literature search of 29 databases was performed on 30 September2008(seeAppendixfor details andthesearchterms used). Wesearched back to 1980, the year of the introduction of DSMIII.22InclusionandexclusioncriteriaOnly publications that included a healthy control groupwereconsideredfor this meta-analysis so that effect sizes couldbecalculated. Drug trials that presented pre-treatment baselinecognitivedatainpatients andincludedahealthycontrol groupwere selected. Publications were excludedif individuals withadiagnosis of schizophrenia and those with another diagnosis(e.g. bipolar disorder withpsychosis) were analysedtogether asone sample. Similarly, publications were also excluded if individuals286Ageatonsetandcognitioninschizophrenia:meta-analysisT. K. Rajji, Z. Ismail andB. H. MulsantBackgroundTherelationshipbetweencognitionandageat onset ofschizophreniaislargelyunknown.AimsTocomparecognitivedeficitsinindividualswithyouth-onsetandlate-onset schizophreniawiththoseinadultswithfirst-episodeschizophrenia.MethodTwenty-ninedatabases(includingEMBASE, MEDLINEandPsycINFO) weresearchedfrom1980to2008. Selectedpublicationshadtoincludehealthycontrolsandanalyseseparatelyindividualsdiagnosedwithschizophreniaor arelateddisorder andindividualswithfirst-episode, youth-onset or late-onset schizophrenia. Descriptiveandcognitivedatawereextractedandthelatter aggregatedinto22cognitivemeasures. Cohenseffect sizerawandweightedmeansof cognitivedeficitsweregeneratedandcomparedinthethreegroups.ResultsIndividualswithyouth-onset andfirst-episodeschizophreniademonstratelargedeficits(meaneffectsize 50.8) onalmostall cognitivemeasures. Individualswithyouth-onsetschizophreniademonstratelarger deficitsthanthosewithfirst-episodeschizophreniaonarithmetic, executivefunction,IQ, psychomotorspeedof processingandverbal memory. Incontrast, thosewithlate-onset schizophreniademonstrateminimal deficitsonarithmetic, digit symbol codingandvocabulary, but larger onesonattention, fluency, globalcognition, IQandvisuospatial construction.ConclusionsIndividualswithyouth-onset schizophreniahaveseverecognitivedeficits, whereasthosewithlate-onsetschizophreniahavesomerelativelypreservedcognitivefunctions. Thisfindingsupportstheviewthat severityof thediseaseprocessisassociatedwithdifferentagesatonset. Inaddition, thecognitivepatternof peoplewithlate-onsetschizophreniasuggeststhat their deficitsarespecificratherthansolelyasaresult of ageingandrelatedfactors.DeclarationofinterestNone.TheBritishJournal of Psychiatry(2009)195, 286293. doi: 10.1192/bjp.bp.108.060723Reviewarticlewith different age at onset (i.e. childhood, adolescence, adultor late-onset) were not analysed separately or if those withadult-onsetschizophreniaincludedpeoplewhowerenotintheirfirst episode. Publications reporting on treatment trials wereexcluded if there were no available baseline data to comparebetween patients and controls. Publications reporting on the samesamplewerecountedasasinglestudy.ClassificationofstudiesIfapublicationmettheabovecriteria,individualswere classifiedinto one of three groups: first-episode schizophrenia, youth-onsetschizophreniaandlate-onset schizophrenia. Tobeclassifiedintoyouth-onset schizophrenia, individuals with schizophrenia hadtohave amaximumage at onset of 19 years. To be classifiedintolate-onset schizophrenia, people with schizophrenia had to have aminimumageatonsetof40years. Foranyotherageatonset, aselectedpublicationwas classifiedintothe first-episode schizo-phreniacategoryprovidedthat thestudyincludedandanalysedseparately those individuals with first-episode schizophrenia.Ageatonsetisdefineddifferentlyindifferentstudies. Itisusedtorefer toageat first behavioural changes (e.g. DeLisi et al23),age at first manifestationof positive symptoms (e.g. White etal16)orageatfirstadmissiontohospital (e.g. Bellgroveetal24).However,thesedifferentstageshavebeenshowntobecorrelatedand typically within 618 months of each other.5,23,25Thus, whenclassifyingastudy, weacceptedthedefinitionof ageatonsetasdescribedineachindividualstudy.ExtractedvariablesAftertheselectionand classificationofapublication, thefollowingvariables wereextractedandrecorded: journal name, title, firstauthor and year of publication. In addition, the followingvariables were recorded for patients and controls separately:numberof participants andmean(s.d.)age. Somestudies onlyprovided an age threshold (e.g. onset before age 14); when mean(s.d.) age at onset was provided, it was recorded and meandurationof illnesswas estimatedbased onthe difference betweenmeanageatonsetandmeanage.Finally,thecognitivedata wererecordedasfollows.CognitivemeasuresThedatafromeachcognitivetest wereextractedandclassifiedintooneof22cognitivemeasures(seeOnlinesupplement). Theselectionof these 22 measures followeda seminal quantitativereviewof cognitivedeficits inschizophrenia.11As notedinthatreview, organizing the myriadof neurocognitive test variablesreported in the literature into a coherent classification [is] a majorchallenge. We adopted the strategy employed in that reviewbecause it accounts for two opposing yet complementaryapproaches. Thefirst approachistoavoidaggregatingcognitivetests andtoreport results of individual tests. This approachissupportedbythefactthatmostindividual testsaremediatedbymultiple cognitive processes and tap into more than one cognitivefunction.Thus, aggregation into asingle cognitive measure couldresult in misleading effect sizes that do not correspond to aspecific cognitive function. This is alsorelevant toour reviewconsidering that indifferent groups (first-episode, youth-onsetandlate-onsetschizophrenia), differentcognitivetestshavebeenused to assess particular cognitive measures. Consequently,contrasting effect sizes that correspond to these cognitivemeasurescouldbeproblematic. Thus, wheneverpossible, resultsof individualtests were reported separately. The second approachis toaggregate cognitive tests into a single measure basedontheoretical or factor-analytical underpinnings. This approachfacilitates the management of data basedonvarious tests thattapintothesamecognitivefunction(e.g. bothCaliforniaVerbalLearning Test and Rey Auditory Verbal Learning Test assess verbalmemory). Italsoaddstotheweightof effectsizesbyincreasingthe number of individual scores contributing to a mean effect size.This is particularly relevant to our review considering therelativelysmallnumberofparticipantsinsomestudies.CalculationandcomparisonofeffectsizesIndividual Cohens effect sizes (di)26were calculated for eachcognitive test as the difference between the means of theschizophrenia and control groups divided by the pooled standarddeviation. In some studies the groups means and standarddeviationswerenotavailableanddi-valueswerecalculatedbasedon the reported inferential statistics.27Effect size rawmeans(d-values) and their 95% confidence intervals (CIs) weregenerated for each cognitive measure based on the individualdi-valuesaggregatedinthismeasure. Finally, effectsizeweightedmeans (weightedd-values)andtheirstandarderrors (s.e.)weregeneratedforeachcognitivemeasurebycalculatingthemeanofindividual di-values weightedbythe inverseof their individualvariance. Weighted means were also generated for age, age at onsetanddurationofillness.WeusedtheQbetweentesttocomparethethreegroupsweightedd-valuesstatistically.28WhenQbetween wassignificant, wecomparedeachof theweightedd-values of eachpair separatelyusingQbetweenwithBonferronis adjustment. Wealso used two difference cut-offs to compare effect sizesqualitatively:0.3and0.6.Followingwell-establishedconventions,aneffectsize