History

P53 immunoexpression as a prognostic marker for human astrocytomas: a meta-analysis and review of the literature

Description
P53 immunoexpression as a prognostic marker for human astrocytomas: a meta-analysis and review of the literature
Categories
Published
of 9
3
Categories
Published
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Similar Documents
Share
Transcript
  LABORATORY INVESTIGATION - HUMAN/ANIMAL TISSUE P53 immunoexpression as a prognostic marker for humanastrocytomas: a meta-analysis and review of the literature Georgia Levidou  • Elias El-Habr  • Angelica A. Saetta  • Christine Bamias  • Klea Katsougiannis  • Efstratios Patsouris  • Penelope Korkolopoulou Received: 24 August 2009/Accepted: 14 April 2010/Published online: 12 May 2010   Springer Science+Business Media, LLC. 2010 Abstract  During the past few decades, researchers havebeen looking for parameters with an impact on the prog-nosis of patients with astrocytic tumors. p53 is one of themost widely investigated molecules in human gliomas. Weaimed to comprehensively review the evidence for theprognostic usefulness of p53 immunohistochemicalexpression in paraffin-embedded tissue specimens fromdiffusely infiltrating astrocytomas. We conducted a sys-tematic review of the PubMed database through December2007 to identify cohort studies that evaluated p53 immu-nohistochemical expression as a prognostic marker forhuman astrocytomas. Estimates of significance wereextracted from association tests and hazard ratios with 95%CI from actuarial curves and Cox regression analyses. Ameta-analysis was performed on the studies that appliedCox models and had adjusted the hazard ratio of p53expression with tumor grade and patients’ age. Wereviewed 44 publications (including 3,627 patients), 14 of which included the estimates (HR and 95% CI) derivedfrom Cox regression. Descriptive analysis showed that mostof the published articles did not contain information onimportant variables such as sex and age (missing from 25%and 11% of studies, respectively), previous treatment, tis-sue-retrieval and follow-up period (56%). The quantitativesynthesis showed that p53 expression is not a significantprognostic factor (combined HR  =  1.034,  P  =  0.531).There was no significant between-study heterogeneity andpublication bias. A second meta-analysis performed only onglioblastomas showed that the overall risk of mortality inthese tumors was  - 0.123 ( - 0.346 to 0.100) and was notstatistically significant. After almost 20 years of research,published evidence does not substantiate the usefulness of p53 immunohistochemical expression as a prognosticmarker in patients with astrocytic neoplasms. Keywords  p53    Immunoexpression   Human astrocytomas    Meta-analysis Introduction Diffusely infiltrating gliomas, and particularly glioblasto-mas, are notorious for their unfavorable prognosis. Fordecades, researchers have been looking for factors allowingthe stratification of patients in prognostic subtypes. Need-less to say, histologic grade remains the most importantfactor influencing prognosis, although grading is notalways sufficient to predict the outcome in individual cases[1]. Thus, there are cases of astrocytic gliomas that unex-pectedly behave less aggressively than other tumors of thesame grade, regardless of the therapeutic intervention. Thisfact suggests the existence of biologic heterogeneity, evenwithin the same histologic grade, that remains to be elu-cidated. Certain clinical parameters, such as patients’ age,tumor location, extent of surgical excision and pre-opera-tive performance status have also been suggested to affectthe outcome [1]. However, research over the last decadeshas indicated that molecular markers involved in pathwaysunderlying the biology of astrocytic tumors also play animportant role as determinants of prognosis [2]. G. Levidou ( & )    E. El-Habr    A. A. Saetta    E. Patsouris   P. KorkolopoulouMedical School, Department of Pathology, Laiko Hospital,National and Kapodistrian University of Athens, 11527 Athens,Greecee-mail: glevidou@yahoo.grG. Levidou    C. Bamias    K. KatsougiannisDepartment of Hygiene, Epidemiology and Medical Statistics,University of Athens, 11527 Athens, Greece  1 3 J Neurooncol (2010) 100:363–371DOI 10.1007/s11060-010-0204-y  P53  is the most frequently involved gene in humancancer. Its protein product has a dual role in the control of cell proliferation and apoptosis [3] either arresting cells inG1 phase to allow replication of damaged DNA or inducingapoptosis when DNA damage is irreversible. Tumor cellsthat have lost p53 function may escape from p53-mediatedcell-cycle arrest and apoptosis and thus display a continuousgrowth of aberrant cells [3]. Overexpression of P53 in thenucleus detected by immunohistochemical techniques iscommonly regarded as a surrogate marker for P53 mutation,and has been one of the most broadly investigated markersin human astrocytomas in the past 15 years [1]. Moleculartechniques have shown that secondary glioblastomas arestrongly associated with p53 mutations, in contrast to pri-mary glioblastomas that are usually marked by epidermalgrowth factor (EGFR) amplification and loss of heteroze-neity (LOH) in chromosome 10 [1].We have performed a systematic review of paperspublished during the past 18 years dealing with changes inP53 immunohistochemical expression and their effect onthe prognosis of diffuse astrocytic gliomas. We also aimedto assess the quality of published studies, to identify factorsthat could affect the assessment of the prognostic role of P53 expression and perform a meta-analysis of availableestimates. Unlike previous published reviews [4, 5], we have included all identified published reports, have asses-sed potential sources of heterogeneity contributing toconflicting results, and have applied quantitative methodsto summarize data. Methods Search strategy and selection criteriaWe systematically reviewed all srcinal articles publishedbetween January 1990 and December 2007 in English,Spanish, German, French, and Italian that analyzed theprognostic role of P53 overexpression in patients withastrocytic tumors. We identified 799 articles from a searchof MEDLINE database using the keywords ‘‘P53’’ and‘‘astrocytoma-in-human’’. The number of studies wasreduced to 77 by limiting the search with the keywords‘‘prognosis’’ or ‘‘prognostic’’ or ‘‘survival’’ and ‘‘immu-nohistochemistry’’ or ‘‘immunohistochemical’’. We alsosearched the reference list of all selected articles andassessed each abstract. Ten studies were selected by thereference list of the aforementioned studies [6–15]. Reviews, non-srcinal articles, and studies analyzing glio-mas other than those of pure astrocytic origin wereexcluded. To avoid duplicate data, we identified articlesthat included the same cohort of patients by reviewing thesimilarity between studies, the country in which the studywas done, the investigators in the study, the source of patients, the recruitment period, and the inclusion criteria.When the same investigators reported results obtained onthe same cohort of patients in several publications, only thelargest series or the series that reported the estimates of their results was included in the analysis. Duplicate reportswere included in the specific analyses only if they applieddifferent antibodies or different immunoreactivity cut-offs,or if they performed different subgroup analyses. A cohortof patients was not included more than once in the sameanalysis.Data extraction and handlingData were extracted in an Access database by two inves-tigators individually (G.L. and E.H.) trained to interpretinformation to ensure homogeneity in data gathering andentry. Complete concordance was reached for all mainvariables assessed in this analysis.The database was designed to ensure that the most rel-evant data on patients (i.e., demographic characteristics,country and period of recruitment, inclusion and exclusioncriteria, study design, tumor characteristics and treatment,endpoint definition, follow-up period, statistical analysis,biologic samples, immunohistochemistry (i.e., antigen-retrieval technique, primary antibodies, staining method,and scoring of immunostaining), and results on the preva-lence of P53 overexpression) were obtained.Statistical analysisWe used two methods to summarize results. Because mostarticles provided only  P  values or statements as to whetherresults were significant (and no other measure of effect), webased our analysis on mortality  P  values (or statements of significance) extracted from association analyses (e.g., chi-square test, Fisher exact test,  t   test, Mann–Whitney test, andlogistic regression), as well as risk estimates and 95% con-fidence intervals (CIs) provided by univariate (i.e., Kaplan–Meier curves and log-rank test), and multivariate survival(i.e., Cox regression) analyses. Unconditional logisticregression models were used to identify the study charac-teristics associated with a significant result ( P \ 0.05).We subsequently performed a meta-analysis by use of multivariate tests. When provided, the adjusted hazard ratio(HR) and 95% CIs for p53 expression were extracted fromthe manuscript. For studies reporting only the total numberof events, the number of patients at risk in each group, andthe statistics and its  P  value, we estimated the appropriateHR using these data, according to the methods described byPakmar and colleagues [16]. In five studies, the researchersreported the raw data, so we performed an analysis in orderto obtain the respective HR estimates [17–21]. In one 364 J Neurooncol (2010) 100:363–371  1 3  study, we received the raw data by regular mail, and thesame calculations were performed [22]. One study alsoreported different estimates for grade III and grade IVcases [6]. In this study, we summarized data using aweighted average of the two reported estimates. We did notuse the reported estimates, in order to avoid reducingbetween-studies heterogeneity, since the same study wouldbe included in the analysis twice. We then used Wolf’smethod to combine risk estimates applying the inverse of variance as the weighting factor. Potential sources of het-erogeneity were investigated through graphical methods,such as the Calbraith plot, whilst a heterogeneity test basedon the Q statistic was carried out. We judged that hetero-geneity was significant when the  P  value was less than0.10. The extent to which the combined risk estimate wasaffected by individual studies was further assessed byconsecutively omitting every study from the meta-analysis.Analyses were done with STATA version 9.0. Results wereregarded as significant at the two-sided  P  value of 0.05.The funding source had no role in the study design, inthe collection, analysis, and interpretation of data, or in thewriting of the report. The corresponding author had fullaccess to all data and had final responsibility to submit thepaper for publication. Results Of 87 studies identified, 8 were excluded because theycontained duplicate data. Moreover, 35 studies wereexcluded either because they were reviews or they did notfulfill the criteria of the meta-analysis protocol. The 44[6–15, 17–50] studies selected comprised 3,627 patients, with sample sizes ranging from 19 to 302 patients (mean82.17, SD 52.92). Twenty-five (56.81%) of these studiesenrolled 80 patients or fewer and only 12 studies (27%)included 100 patients or more. Twenty (45%) of the studiesassessed patients from Europe, 11 (24%) from the USA, 10(22%) from Asia, and 3 (7%) from the rest of the world.The size of the studies did not seem to vary according tothe location of the study ( P  =  0.14).The median patient recruitment period was 9 years(range 2–18). Mean age of patients was 44.5 years (SD14.54) and 60% of patients (range 30–84%) were men. Allstudies selected patients retrospectively. Sixteen (36%)studies excluded patients from their analyses, mainlybecause of absence of tissue or unavailable follow-up data.None of the studies compared the characteristics of excluded patients with those of the srcinal cohort in orderto assess whether absence of tissue was associated withhistologic grade.One (2%) article included only pilocytic astrocytomas,7 (16%) investigated only diffuse astrocytomas (grade 2),5 (11%) only anaplastic astrocytomas (grade III), 14 (31%)only glioblastomas (grade 4), and the other 18 (40%)included more than one tumor grade. A total of 1,975(54.45%) patients had glioblastomas. Twenty-five (58%)studies reported detailed information on treatment charac-teristics, 7 (16%) studies provided limited information inthis regard, whereas the remaining 11 studies (26%) did notdescribe the treatment strategy used.Table 1 shows the number of studies that did not pro-vide information on important variables. Basic sociode-mographic information, such as sex and age was missingfrom 25 and 11% of studies, respectively. Pathologiccharacteristics of tumors were well reported, whereas otherimportant variables (i.e., previous treatment and tissue-retrieval period) were not. Furthermore, 25 of 44 did notstate the follow-up period.The immunohistochemical techniques used variedwidely among studies. The most commonly used antibod-ies were DO-7 ( n  =  20, 44%), PAb1801 ( n  =  9, 20%) andDO-1 ( n  =  8, 18%); the other antibodies used were CM-1,Pab1620, M 7001, PAb 204 and BP53-12. Four (9%)articles did not define the primary antibody used. A widerange of dilutions was used (from 1/20 to 1/64,000). Of the31 (69%) studies that reported the immunohistochemicalassay used, 29 (94%) used avidin–biotin peroxidase.The definition of positive P53 staining (i.e., nuclearoverexpression) also varied widely among studies. The cut-off values used by various investigators ranged from 1 to50%. In particular, in 25 studies (56.8%), the cut-off valueused to define positive staining was  * 1%, in only threestudies was the respective value higher than 25%, whereasin the remaining 13 studies, the cut-off value ranged Table 1  Studies without information on selected important variablesVariable Lacking informationDescriptive dataAge 5/44 (11%)Gender 11/44 (25%)Inclusion criteria 37/44 (86%)Exclusion criteria 28/44 (64%)Follow-up period 25/44 (57%)Recruitment time 11/44 (25%)p53 immunihistochemical analysisTumour area selection 17/44 (39%)Primary antibody 1/45 (2%)Antibody dilution 1/44 (2%)Antigen retrieval technique 2/44 (4%)Immunohistochemistry (IHC) 14/44 (32%)Definition of positive staining 1/44 (2%)Double assessment of IHC results 39/44 (89%)Blind assessment of IHC results 38/44 (86%)J Neurooncol (2010) 100:363–371 365  1 3  between 3 and 10%. These three studies, however, were noteligible for inclusion in meta-analysis, a fact that seems toweaken a possible bias made by this between-study vari-ation. Morever, only in five (11%) studies was the assess-ment of p53 immunopositivity performed by more than oneobserver, and in four studies a concordance analysis forP53 positivity among the observers was carried out. Nostudy reported Kappa coefficients .  Immunohistochemicalassessment was made by investigators who were unawareof the clinical information in 6 (13%) studies.Eight (18%) studies based their results on associationtests and did not apply survival analysis. Univariate sur-vival analysis was applied in 36 (86%) of the studies, andCox regression in 31 (74%). Important data regarding thecorrelation of p53 with survival were reported in 15 studiesthat performed multivariate analysis. The mean follow-upfor all studies was 35.93 months (SD 23.809).Thirty (97%) studies reported the cofactors used in themultivariate models, which unfortunately varied widely,even for a given endpoint. The most common cofactors inthe 31 studies that used multivariate analysis to assess therisk of mortality (one study did not state the factorsadjusted for) were age ( n  =  27), Karnofsky performancescore ( n  =  9) and Ki-67 index ( n  =  14). Histologic gradewas used in 8 studies from the 11 that assessed a multi-graded cohort.Table 2 shows the significances for the comparison of P53 overexpression with prognosis of human astrocytomasas reported in the 15 studies for whom the estimates formultivariate survival analysis were available. All primaryantibodies used in the evaluation of the expression of p53were represented in this cohort of studies. The proportion of each histologic grade was the following: 20% grade II, 17%grade III, and 63% grade IV. No clear association of P53overexpression with mortality was noted in 12 of 19 studies(data not shown) that used univariate tests, or in 12 of 15studies thatused multivariatetestsand reported their results.Association and univariate tests reported significant resultsmore frequently than multivariate survival analysis did.Several characteristics of the design of the studies couldaccount for the wide heterogeneity in results, includingsample size, homogeneity of the series (e.g, stage andgrade), geographical area where the research was done,year of publication, length of recruitment period, inclusioncriteria, previous treatment, sample storage, primary anti-body and dilution, antigen-retrieval technique, cut-off value, endpoint definition, follow-up period, statisticalstrategy, and adjustment for cofactors.We therefore investigated which factors independentlyaffected results of survival analyses by using both stratifiedand multivariate unconditional logistic regression modelsfor every endpoint. The dependent variable was a  P  valueor less than or equal to 0.05. Year of publication, absenceof information on inclusion criteria and use of multivariateanalysis were not significantly associated with studies thatfound an association between P53 changes and survivaleither in univariate or in multivariate analysis. However,the use of DO-1 and DO-7 primary antibodies seemed to bemarginally correlated with the presence of an associationbetween P53 and survival ( P  =  0.068). Table 2  Characteristics of studies included in the meta-analysis for the prognostic role of p53 expression in patients with astrocytic tumoursFirst author Year of publication n  Grade Log(HR) SE[log(HR)]  P  value ReferenceKoshunov A. 1999 168 4  - 0.0943 0.1638 0.54 33Layfield L.G. 2006 32 4  - 0.2487 0.1927 0.526 20Haapasalo H. 1999 78 1 0.5417 0.1073 0.985 26Jaros E. 1992 43 1, 2, 3, 4 0.1909 0.1599 0.633 18Korkolopoulou P. 1997 51 2, 3, 4 0.0339 0.1214  0.0217  46Kyritsis A.P. 1995 182 3  - 0.5798 0.3409  0.015  6Kyritsis A.P. 4 0.0295 0.1155 0.80Nakamura M. 1997 50 3 0.3987 0.43654 0.320 45Pardo F.S. 2004 74 2, 3, 4 0.641 0.2949  0.0426  28Rainov N.G. 1997 107 3, 4  - 0.67 0.32  0.04  44Saito T. 2006 50 4 0.2844 0.3135 0.455 24Vital A. 1998 100 2 0.3911 0.2753 0.158 22Drach L.M. 1996 21 2, 3 0.01425 0.8358 0.986 17Muhammad A.K.M. 1997 19 2, 3, 4 0.6399 0.3955 0.309 21Cunninham J.M. 1997 120 2, 3, 4 0.0086 0.2126 0.77 9DeMasters B.K. 2006 38 4  - 0.79604 0.6066 0.184 19Bold defines the articles that reported statistically significant results ( P  value \ 0.05)366 J Neurooncol (2010) 100:363–371  1 3  Table 2 shows the characteristics of the qualified studiesfor inclusion in meta-analysis, which overall did not showconclusive results. Our meta-analysis included only thoseestimates (HR and 95% CIs) derived from Cox regressionmodels. The study by Haapasalo and colleagues [26] wasexcluded from the analysis since it investigated only pilo-cytic astrocytomas. Moreover, Cunningham and colleagues[9] performed a double assessment of p53 expression,using two different primary antibodies, and then adjustedtwo different models in order to explore the correlation of p53 expression with survival. In our analysis, we use onlythe assessment of staining with DO-7 antibody since it wasperformed in the whole cohort presented in the article.Thus, 14 publications were finally included in the meta-analysis.Cochran’s statistical test for heterogeneity did notestablish the presence of heterogeneity between studies( P  =  0.169). Graphical exploration for heterogeneity wasalso negative, although the study by Pardo and colleagues[28] seems to be marginally outside the 95% CIs of thecombined estimate (Fig. 1a). We therefore used the fixedeffects model. The overall risk of mortality was 1.034( - 0.073 to 0.141) and was not statistically significant(Fig. 2a). Begg’s and Egger’s statistical tests and sche-matic representations for the assessment of publication biaswere also not significant ( P  =  0.502 for Begg’s and P  =  0.214 for Egger’s). Sensitivity analysis showed thatthe most influential studies were those of Kyritsis andcolleagues [6] and Layfield and colleagues [20] (Fig. 3). Since the majority of the studies (14) included onlyglioblastomas, we performed a separate meta-analysis forthese studies. Unfortunately, only 4 studies were eligiblefor inclusion in this meta-analysis (19, 20, 24, 32).Graphical and statistical assessemnt of heterogeneity werenegative ( P  =  0.337; Fig. 1b). Fixed effects model showedthat the overall risk of mortality in these tumors was - 0.123 ( - 0.346 to 0.100) and was not statistically sig-nificant (Fig. 2b). The assessments of publication bias werealso not significant ( P  =  0.734 for Begg’s test and P  =  0.802 for Egger’s test). Fig. 1  Gallbraith plot for the meta-analysis of the prognostic role of p53 expression in human astrocytomas (14 studies included) Fig. 2  Hazard ratios and 95% CI of studies included in meta-analysis of the prognostic role of p53 expression in human astrocytomasJ Neurooncol (2010) 100:363–371 367  1 3
Search
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x