SEARCH · PubMed Health

Results for “Classification”

Explore indexed PubMed citations for clinical trials, systematic reviews and public health research. Read source abstracts and follow each citation to its original PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 577 records · Page 32Linked to original sources

PubMed1998

Biologic classification as an alternative to anatomic staging for clinically localized prostate cancer: a proposal based on patients treated with external beam radiotherapy.

OBJECTIVES: The prognostic significance of clinical stage in patients with prostate cancer who are treated with external beam radiotherapy is unclear. This study evaluates multiple pretreatment factors, including clinical stage, to determine which are the best prognostic factors, and develops a classification system based on these factors. METHODS: All 249 evaluable patients with clinically localized adenocarcinoma of the prostate treated with definitive conformal external beam radiotherapy without androgen deprivation at our institution between 1989 and 1993 were analyzed. Clinical stage, serum PSA level, Gleason score, race, and history of transurethral resection of the prostate (TURP) were evaluated for their ability to predict biochemical disease-free survival (BDFS). Factors predictive of BDFS were then used to construct a classification system. The classification system was then analyzed for its ability to predict BDFS, distant metastases, local recurrence, and clinical disease free survival in univariate and multivariate analyses. Median follow-up was 27 months. RESULTS: Gleason score and PSA predicted BDFS in multivariate analysis (both P <0.0001), whereas clinical stage, race, and history of a TURP did not. These two biologic factors were combined into a four-level classification system. This classification system was analyzed together with Gleason score and PSA and was found to be the only predictor of BDFS on multivariate analysis (P <0.0001). In addition, this classification system was the only predictor of distant metastases in multivariate analysis (P <0.0001). CONCLUSIONS: The classification system derived herein based on the biologic factors of Gleason score and serum PSA levels is the sole predictor of distant metastases and biochemical recurrence for patients treated with definitive conformal external beam radiotherapy for clinically localized prostate cancer. This classification system may be useful when comparing competing therapies and stratifying patients in clinical trials, but requires validation from other institutions and other therapies prior to its widespread use.

Adenocarcinoma↗

PubMed1993

Classification of endometriosis.

A perfect endometriosis classification should be a common language and an expert system which helps the gynaecologist to decide the treatment of each patient. The staging of endometriosis is not a new idea; the first classification based on histologic criteria was presented in 1941. Since this initial attempt, several classifications have been proposed. These various systems are reviewed in terms of their advantages and defects. The revised American Fertility Society (AFS) endometriosis classification system is now accepted worldwide as the endometriosis international language. Laparoscopic staging and measurement techniques are presented and discussed. In recent reports, changes to the revised AFS classification have been proposed, including a better description of atypical and deep infiltrating peritoneal implants, and a stage V for patients with bilateral, extensive, dense adhesions. Despite its well-known advantages, the revised AFS classification cannot be used as a satisfactory expert system. A better understanding of endometriosis is required to improve the present system. Scientifically based scores for each lesion and a marker for disease 'activity' will be fundamental to a classification which will act as a valuable expert system and become the endometriosis classification of the twenty-first century.

Endometriosis↗

PubMed2003

Comments on the 2001 WHO proposal for the classification of haematopoietic neoplasms.

In the preface, the World Health Organization (WHO) classification vows to offer pathologists, oncologists and geneticists worldwide a system of classification for human neoplasms based on histopathological and genetic features. Standardization of nomenclature and agreed-upon criteria for definition of the various types of cancer are felt to be a prerequisite for progress in clinical oncology, multicentre therapy trials and comparative studies in different countries. In fact, the WHO effort represents the first worldwide comprehensive consensus classification of the haematological malignancies. Consensus was reached among a subgroup of investigators, carefully selected for their experience and contributions to existing classifications. In the present climate of daily new discoveries that yield a constant stream of fascinating insights into the biology of leukaemias and lymphomas and, above all, resulting in an explosion of potential therapeutic targets, the WHO system has taken the stand of compiling established classification approaches and providing order to known facts. This furnishes an essential skeleton upon which to build in the future. The WHO committee decided that sorting neoplasms according to prognosis was neither practical nor necessary and could be misleading. While justifiable at the present time, it is important to realize that the classifications of the haematological malignancies are a moving target and that the trend is to move away from currently accepted gold standards, such as morphological evaluations, in favour of genetic characterizations, especially those with therapeutic relevance. The goal of this chapter is to fill in some gaps that, as per the author's opinion, exist in the WHO classification, predominantly, where it concerns the role of immunophenotyping as a complementary discipline for genotyping through its potential to generate surrogate marker profiles for molecular lesions. By introducing some state-of-the-art classification modalities, some of which are still awaiting confirmation, this chapter also aims to spark excitement and provide a glimpse at the future.

Genotype↗

PubMed1981

Individual differences in the classification of stimuli by dimensions.

In two experiments, adults differing in general mental abilities ("intelligence") were given multidimensional classification tasks: aural and visual free-classification tasks that allowed grouping of triads or trends by conflicting criteria of dimensional identity or overall similarity, and a speeded classification task that required the filtering of variation in an irrelevant dimension. Four principal results are reported: (a) Individual differences in the tendency to use similarity criteria in free classification were systematic across modalities and stimulus dimensions; (b) filtering performance, measured as the effect of irrelevant variation on speeded classification, did not correlate with free-classification performance (percent similarity responses); (c) the tendency to use similarity criteria in free classification was not correlated with intelligence; and (d) the amount of interference in speeded classification was correlated negatively with intelligence. The results taken together suggest that there are two dimensions of individual differences in perceptual analysis, one having to do with ability and the other with style or preference. This suggestion was supported by an additional analysis. The results also raise questions about the convergence of measures of multidimensional stimulus processing.

Discrimination Learning↗

PubMed2001

Classification of headaches.

It was not until 1962 that the Ad-Hoc Committee of the National Institute of Health first published a classification of headache syndromes by brief glossary definitions. The general disadvantage of such glossary definitions is that they require subjective interpretation. Therefore under the chairmanship of Prof. Jes Olesen, Copenhagen, the International Headache Society published in 1988 on the basis of empirical findings a first ever headache classification using operationalized criteria. The headache classification of the International Headache Society was immediately translated into the world's major languages and was adopted by all national headache societies represented in the International Headache Society, the World Health Organisation and the World Federation of Neurology. The new classification proved so successful and enjoyed such rapid international acceptance that no revision was undertaken until 1999. The second edition, again under the chairmanship of Prof. Jes Olesen, will probably be completed in 2002. The classification produced such a high degree of inspiration and motivation of pathophysiological and epidemiological research work that knowledge in the field of headache has displayed growth unparalleled in any other field of neurological research. This development was made possible by the determined work of the Chairman of the Headache Classification Committee, Prof. Jes Olesen. He succeeded in bringing together international researchers, motivating them and jointly turning the current fund of knowledge into a evidence-based classification. Prof. Jes Olesen thus performed the decisive pioneering work for all those who have to do with headaches-patients, doctors and scientists. The IHS classification is the most frequently cited text and one of the most important milestones in the history of the scientific study of headaches.

Headache↗

PubMed2002

Comparison of two T-classification systems for sino-nasal carcinoma.

It is often difficult to determine the actual site of origin of tumours originating in the sino-nasal region, and a uniform classification system that covers all tumours in this area is warranted. A retrospective series of 165 consecutive patients with sino-nasal carcinoma, treated and followed at the Aarhus University Hospital between 1963 and 1991, was evaluated and T-staged according to the Lederman classification. The 80 maxillary antrum carcinomas were also staged according to the UICC 1997 system. In univariate analysis, the UICC T-classification was prognostic for locoregional tumour control and disease-specific survival. However, when adjusted for covariates (gender and nodal involvement) in a multivariate analysis, the UICC classification was not a significant independent prognostic parameter. In contrast, the Lederman T-classification was prognostic both in univariate and multivariate analysis. The Lederman T-classification was more prognostic for locoregional control and disease-specific survival than the UICC TNM classification. In addition, the Lederman classification is easy to use and has a broader applicability as it covers all sites in the sino-nasal area.

Adenocarcinoma↗

PubMed1993

Testing the validity of a prognostic classification in patients with surgically optimal ovarian carcinoma: a 15-year review.

A retrospectively designed classification using stage, residuum and a variable which combines prognostic information from both grade and histology (histology-grade variable) has been used at our institution to predict prognosis, and choose therapy in patients with ovarian carcinoma, stages I-III having no or small residuum. In this study, multivariate analysis of prognostic factors were performed over two time periods: Group 1 (1971-1978), contains the patients from which the original classification was derived, and Group 2 (1979-1985), contains a different cohort of patients who are used to test the validity and reproducibility of the original classification. Multivariate analysis showed that the prognostic significance of two variables changed over the two study periods: tumor grade, and residuum. It was found that in the ideal combination of grade and histologic type, when used in conjunction with stage and residuum in a prognostic classification, was unique to each patient cohort. Because of these changes, new and more accurate prognostic classifications were derived for Group 2. However, when all classifications were examined, (including the original), the differences in their ability to stratify patients into risk categories was negligible, and there was no major advantage to using one classification over another for clinical applications. Thus, the retrospectively derived prognostic classification using grade, instead of a combined histology-grade variable, in conjunction with the other significant prognostic factors (stage and residuum), is preferred for prospective application, and for its simplicity.

Journal Article↗

PubMed2002

Comparison of causes of death using HEMO Study and HCFA end-stage renal disease death notification classification systems. The National Institutes of Health-funded Hemodialysis. Health Care Financing Administration.

Few data are available on the accuracy of death classification in patients with end-stage renal disease (ESRD). The National Institutes of Health-funded Hemodialysis (HEMO) Study allows the opportunity to compare cause of death recorded on the Health Care Financing Administration (HCFA) Death Notification Form 2746 with death classified by the HEMO Study. The HEMO Study cause of death is determined by trained HEMO Study Outcome Review Committee physicians. In this interim analysis, there were 220 deaths coded by both classification systems. Using the HEMO Study classification system, the most common cause of death was ischemic heart disease (20.4%), followed by arrhythmia and conduction problems (10.4%), cerebrovascular disease (8.6%), and non-access-related infections (7.7%). Using the HEMO Study final death classification as the reference standard, most differences in the two classification systems were related to coding of heart disease. Sensitivity for the HCFA classification ranged from 9.1% for congestive heart failure to 91.7% for malignancy, whereas specificity values were all greater than 78%. Positive predictive values ranged from 11.8% for other heart disease and conditions to 100% for malignancy and hepatobiliary disease, whereas negative predictive values were all greater than 85%. The kappa statistic between the two death classification systems ranged from 0.12 for congestive heart failure to 0.95 for malignancy. Studies using death classification from the HCFA ESRD death notification form for deaths secondary to either cardiovascular diseases or unknown causes should be interpreted cautiously.

Aged↗

PubMedDate not supplied

[The problem of comparability of results of the treatment of congenital hip dislocation--exemplified by the Severin scheme and the classification system of the Work Group Hip Dysplasia].

The comparison of the results obtained by conservative or operative treatment of hip dysplasia shows that the consideration of absolute roentgenologic hip parameters is of little use. Classification principles for numerous hip parameters have been developed for this reason by Tönnis. In the Angloamerican and Scandinavian countries, on the other hand, it is predominantly the Severin classification that is employed for an evaluation of the methods of treatment. We examined in our study the differences in the evaluation of the results obtained in 117 hip joints treated by pelvic osteotomy according to Salter by application of the AKH and the Severin classification scheme. There were evident differences in all groups of these classifications. In the application of the Severin scheme the groups I, II and III contained 9.5% less, 24.8% more, and 18.8% less hip joints, respectively than in the respective groups of the Tönnis classification. In the groups IV, V and VI of the Severin classification there were altogether 3.5% less hip joints than in the respective groups of the Tönnis classification. Our study thus showed that a comparison of the operative results of hip dysplasia as presented in the Angloamerican and Scandinavian literature and those presented in the German literature is not possible. We conclude from this result that an evaluation according to a unified classification scheme would be most desirable.

Adolescent↗

PubMed2003

World Health Organization classification of hematopoietic and lymphoid tissues: implications for dermatology.

The new World Health Organization (WHO) classification of hematopoietic and lymphoid malignancy represents the first worldwide consensus document on the classification of lymphoma/leukemia. Its aim is to supercede all existing classification systems, including the Revised European American Classification (REAL), and the organ-based classification of cutaneous lymphoma proposed in 1997 by the European Organization for Research and Treatment of Cancer (EORTC) cutaneous lymphoma project group. This article compares the REAL, EORTC, and WHO classifications with particular reference to cutaneous lymphoma. It identifies those entities that have been adopted from the EORTC classification, and illustrates important new entities in the category of cytotoxic T-/NK-cell lymphomas that characteristically present in the skin or subcutaneous tissue. However, WHO is not an organ-based classification and some categories (eg follicular B-cell lymphoma and peripheral T-cell lymphoma) include both cutaneous and systemic diseases, and these may have very different prognoses. It is, therefore, important that cancer registries are capable of recording sites of involvement for both B- and T-cell lymphomas so that the incidence and prognosis of cutaneous lymphomas can be established from national data sets in future years.

Hematologic Neoplasms↗

PubMed2001

Reliability of the AO/ASIF classification for pertrochanteric femoral fractures.

20 radiographs of pertrochanteric femoral fractures were classified as to fracture "group" and "sub-group" according to the AO/ASIF Fracture Classification (type 31A) by 15 observers. 3 months later, the same radiographs were reviewed by the same observers. Mean agreement of the observers with the final consensus ranged from 53% (with subgroup classification) to 81% (without subgroup). The mean kappa value for interobserver reliability was 0.33 and 0.34 for classification with subgroup in both observer sessions, respectively. Omission of the subgroup classification resulted in better mean kappa values (0.67 and 0.63, respectively). Mean intraobserver reliability was 0.48 in the fracture "subgroup" and 0.78 in the "group" classification. In conclusion, the results show that the AO/ASIF classification for pertrochanteric fractures is reliable for fracture subgroups 31A1, A2 or A3. The group classification should be used to compare scientific data and determine the best treatment. Further classification of fracture subgroups leads to poor reproducibility of results.

Femoral Fractures↗

PubMed2002

Inter-observer reliability of radiographic classifications and measurements in the assessment of Perthes' disease.

We evaluated the inter-observer agreement of radiographic methods when evaluating patients with Perthes' disease. The radiographs were assessed at the time of diagnosis and at the 1-year follow-up by local orthopaedic surgeons (O) and 2 experienced pediatric orthopedic surgeons (TT and SS). The Catterall, Salter-Thompson, and Herring lateral pillar classifications were compared, and the femoral head coverage (FHC), center-edge angle (CE-angle), and articulo-trochanteric distance (ATD) were measured in the affected and normal hips. On the primary evaluation, the lateral pillar and Salter-Thompson classifications had a higher level of agreement among the observers than the Catterall classification, but none of the classifications showed good agreement (weighted kappa values between O and SS 0.56, 0.54, 0.49, respectively). Combining Catterall groups 1 and 2 into one group, and groups 3 and 4 into another resulted in better agreement (kappa 0.55) than with the original 4-group system. The agreement was also better (kappa 0.62-0.70) between experienced than between less experienced examiners for all classifications. The femoral head coverage was a more reliable and accurate measure than the CE-angle for quantifying the acetabular covering of the femoral head, as indicated by higher intraclass correlation coefficients (ICC) and smaller inter-observer differences. The ATD showed good agreement in all comparisons and had low interobserver differences. We conclude that all classifications of femoral head involvement are adequate in clinical work if the radiographic assessment is done by experienced examiners. When they are less experienced examiners, a 2-group classification or the lateral pillar classification is more reliable. For evaluation of containment of the femoral head, FHC is more appropriate than the CE-angle.

Body Weights and Measures↗

PubMed1998

Lack of precision in neonatal death classifications based on the underlying causes of death stated on death certificates.

Large-scale analyses of causes of neonatal deaths are usually based on death-certificate information. A new computer-based method has been introduced to define the cause of stillbirths and neonatal deaths in large amounts of material and to classify them according to two different models [Wigglesworth and Neonatal and Intrauterine death Classification according to (a)Etiology (NICE)]. The method is based on a combination of detailed information from health care registries and the death-certificate information. The present study aimed to compare these two classification models with a previously published method based solely on death certificate information [International Collaborative Effort (ICE)]. The study population comprised 2378 neonatal deaths in Sweden between 1987 and 1992. Cross-tabulation was made between the ICE classification and the other two classification models. In addition, case examples are presented in detail, exemplifying how classification errors arose. The ICE classification gives a rather low precision, notably for two important causes of death: asphyxia and immaturity. Among 328 infants dying from asphyxia according to computerized Wigglesworth classification, ICE classified 59% as asphyxia and 22% were labelled immaturity. When ICE classified the deaths as due to asphyxia, this was verified in only 50%. Among 792 infants dying from immaturity according to computerized Wigglesworth classification, 64% were classified as such by ICE. The findings cast doubts on the results of studies based exclusively on death-certificate information. Whenever possible in the analysis of neonatal deaths, death-certificate information should be supplemented with more detailed data. The computer-based method introduced here makes such analyses possible for large databases.

Cause of Death↗

PubMed2003

Simultaneous gene clustering and subset selection for sample classification via MDL.

MOTIVATION: The microarray technology allows for the simultaneous monitoring of thousands of genes for each sample. The high-dimensional gene expression data can be used to study similarities of gene expression profiles across different samples to form a gene clustering. The clusters may be indicative of genetic pathways. Parallel to gene clustering is the important application of sample classification based on all or selected gene expressions. The gene clustering and sample classification are often undertaken separately, or in a directional manner (one as an aid for the other). However, such separation of these two tasks may occlude informative structure in the data. Here we present an algorithm for the simultaneous clustering of genes and subset selection of gene clusters for sample classification. We develop a new model selection criterion based on Rissanen's MDL (minimum description length) principle. For the first time, an MDL code length is given for both explanatory variables (genes) and response variables (sample class labels). The final output of the proposed algorithm is a sparse and interpretable classification rule based on cluster centroids or the closest genes to the centroids. RESULTS: Our algorithm for simultaneous gene clustering and subset selection for classification is applied to three publicly available data sets. For all three data sets, we obtain sparse and interpretable classification models based on centroids of clusters. At the same time, these models give competitive test error rates as the best reported methods. Compared with classification models based on single gene selections, our rules are stable in the sense that the number of clusters has a small variability and the centroids of the clusters are well correlated (or consistent) across different cross validation samples. We also discuss models where the centroids of clusters are replaced with the genes closest to the centroids. These models show comparable test error rates to models based on single gene selection, but are more sparse as well as more stable. Moreover, we comment on how the inclusion of a classification criterion affects the gene clustering, bringing out class informative structure in the data. AVAILABILITY: The methods presented in this paper have been implemented in the R language. The source code is available from the first author.

Algorithms↗

PubMed2005

Robust classification modeling on microarray data using misclassification penalized posterior.

MOTIVATION: Genome-wide microarray data are often used in challenging classification problems of clinically relevant subtypes of human diseases. However, the identification of a parsimonious robust prediction model that performs consistently well on future independent data has not been successful due to the biased model selection from an extremely large number of candidate models during the classification model search and construction. Furthermore, common criteria of prediction model performance, such as classification error rates, do not provide a sensitive measure for evaluating performance of such astronomic competing models. Also, even though several different classification approaches have been utilized to tackle such classification problems, no direct comparison on these methods have been made. RESULTS: We introduce a novel measure for assessing the performance of a prediction model, the misclassification-penalized posterior (MiPP), the sum of the posterior classification probabilities penalized by the number of incorrectly classified samples. Using MiPP, we implement a forward step-wise cross-validated procedure to find our optimal prediction models with different numbers of features on a training set. Our final robust classification model and its dimension are determined based on a completely independent test dataset. This MiPP-based classification modeling approach enables us to identify the most parsimonious robust prediction models only with two or three features on well-known microarray datasets. These models show superior performance to other models in the literature that often have more than 40-100 features in their model construction. AVAILABILITY: Our MiPP software program is available at the Bioconductor website (http://www.bioconductor.org).

Algorithms↗

PubMed2006

Reliable gene signatures for microarray classification: assessment of stability and performance.

MOTIVATION: Two important questions for the analysis of gene expression measurements from different sample classes are (1) how to classify samples and (2) how to identify meaningful gene signatures (ranked gene lists) exhibiting the differences between classes and sample subsets. Solutions to both questions have immediate biological and biomedical applications. To achieve optimal classification performance, a suitable combination of classifier and gene selection method needs to be specifically selected for a given dataset. The selected gene signatures can be unstable and the resulting classification accuracy unreliable, particularly when considering different subsets of samples. Both unstable gene signatures and overestimated classification accuracy can impair biological conclusions. METHODS: We address these two issues by repeatedly evaluating the classification performance of all models, i.e. pairwise combinations of various gene selection and classification methods, for random subsets of arrays (sampling). A model score is used to select the most appropriate model for the given dataset. Consensus gene signatures are constructed by extracting those genes frequently selected over many samplings. Sampling additionally permits measurement of the stability of the classification performance for each model, which serves as a measure of model reliability. RESULTS: We analyzed a large gene expression dataset with 78 measurements of four different cartilage sample classes. Classifiers trained on subsets of measurements frequently produce models with highly variable performance. Our approach provides reliable classification performance estimates via sampling. In addition to reliable classification performance, we determined stable consensus signatures (i.e. gene lists) for sample classes. Manual literature screening showed that these genes are highly relevant to our gene expression experiment with osteoarthritic cartilage. We compared our approach to others based on a publicly available dataset on breast cancer. AVAILABILITY: R package at http://www.bio.ifi.lmu.de/~davis/edaprakt

Algorithms↗

PubMed2006

A Protein Classification Benchmark collection for machine learning.

Protein classification by machine learning algorithms is now widely used in structural and functional annotation of proteins. The Protein Classification Benchmark collection (http://hydra.icgeb.trieste.it/benchmark) was created in order to provide standard datasets on which the performance of machine learning methods can be compared. It is primarily meant for method developers and users interested in comparing methods under standardized conditions. The collection contains datasets of sequences and structures, and each set is subdivided into positive/negative, training/test sets in several ways. There is a total of 6405 classification tasks, 3297 on protein sequences, 3095 on protein structures and 10 on protein coding regions in DNA. Typical tasks include the classification of structural domains in the SCOP and CATH databases based on their sequences or structures, as well as various functional and taxonomic classification problems. In the case of hierarchical classification schemes, the classification tasks can be defined at various levels of the hierarchy (such as classes, folds, superfamilies, etc.). For each dataset there are distance matrices available that contain all vs. all comparison of the data, based on various sequence or structure comparison methods, as well as a set of classification performance measures computed with various classifier algorithms.

Algorithms↗

PubMed1999

An evaluation of the Banff classification of early renal allograft biopsies and correlation with outcome.

BACKGROUND: The Banff classification for assessment of renal allograft biopsies was introduced as a standardized international classification of renal allograft pathology and acute rejection. Subsequent debate and evaluation studies have attempted to develop and refine the classification. A recent alternative classification, known as the National Institutes of Health Collaborative Clinical Trials in Transplantation (NIH-CCTT) classification, proposed three distinct types of acute rejection. The 1997 Fourth Banff meeting appeared to move towards a consensus for describing transplant biopsies, which incorporated both approaches. Patients who received a renal allograft at the Oxford Transplant Centre were managed by a combination of protocol and clinically indicated biopsies. We have undertaken a retrospective analysis of the biopsies correlated with the clinical outcome to test the prognostic value of the original Banff (Banff 93-95) and NIH-CCTT classifications. METHODS: Three hundred and eighty-two patients received renal allografts between May 1985 and December 1989, and were immunosuppressed using a standard protocol of cyclosporine, azathioprine and steroid. Adequate 5-year follow-up data were available on 351 patients, and of these, 293 had at least one satisfactory biopsy taken between days 2 and 35 after transplantation, the latter patients forming the study group. The D2-35 biopsies taken from these patients, which were not originally reported according to the Banff classification, were re-examined and classified according to the Banff 93-95 protocols. For each patient the biopsy found to be the most severely abnormal was selected, and the Banff and NIH-CCTT grading compared with the clinical outcome. RESULTS: Seven hundred and forty-three biopsies taken from 293 patients between days 2 and 35 after transplantation were examined and the patients categorized on the basis of the 'worst' Banff grading as follows. Normal or non-rejection, 20%; borderline, 34%; acute rejection grade I (AR I), 18%; AR IIA, 6%; AR IIB, 14%; AR III, 1%; AR IIIC, 3%; widespread necrosis 3%. The clinical outcome for the last two groups combined was very poor with 18% of grafts functioning at 3 months and 6% at 5 years. The other groups with vascular rejection (AR IIB and AR III) had an intermediate outcome, graft survival being 78% at 3 months and 61% at 5 years. The remaining four groups (normal, borderline, cellular AR I and AR IIA) had the best outcome: graft survival 95% at 3 months and 78% at 5 years with virtually no difference between the four groups. Three forms of acute rejection, namely tubulo-interstitial, vascular and transmural vascular, were identified, but only the latter two categories were associated with a poor outcome. CONCLUSIONS: The eight sub-categories of the Banff classification of renal allograft biopsies are associated with three different prognoses with respect to graft survival in the medium term. These three prognostic groups correspond to the three NIH-CCTT types. The data provide support for the consensus developed at Banff 97 separating tubulo-interstitial, vascular and transmural vascular rejection (types I, II and III acute rejection).

Evaluation Studies as Topic↗