| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 808.56 KB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
No contexto dos serviços de sangue, a segurança pós‐transfusão do recetor de sangue é considerada um tópico maior.
Considerando a segurança dos recetores de sangue, estes serviços deverão testar um conjunto de agentes transmissíveis com
prevalência significante na população. O papel dos resultados dos testes de rastreio na validação das dádivas de sangue é crítico
para a garantia da segurança pós‐transfusão. Nos países da União Europeia, o rastreio é requerido pela diretiva 2001/83/EC e
nos EUA pelas “normas para bancos de sangue e serviços de transfusão”.
Dada a incerteza associada aos testes de rastreio, um resultado reportado poderá ser falso. Se um resultado falso positivo não
tem consequência para a segurança pós‐transfusão, já um resultado falso negativo tem um grande impacto já que existe uma
probabilidade elevada de o recetor do componente sanguíneo ser infetado. Consequentemente, os laboratórios de rastreio
devem considerar a avaliação da incerteza de medição nos esquemas de controlo de qualidade. Contudo, os requisitos da União
Europeia ou dos EUA relacionados com incerteza são genéricos.
Esta tese reúne um conjunto de artigos com discussões acerca de métodos para a determinação da incerteza da medição em
imunoensaios de rastreio, assim como de abordagens complementares e modelos de exatidão diagnóstica. A utilização do Guia
para a Expressão da Incerteza na Medição (GUM) não é sistemática nos laboratórios de rastreio dos serviços de sangue, tal
como noutros laboratórios clínicos. Contudo, têm sido usados nos serviços de sangue outras estimativas de incerteza usando
métodos estatísticos. Assim, a discussão reúne, não só modelos de acordo com os princípios do GUM focados na incerteza de
resultados próximos ou iguais ao “cutoff” (ponto de decisão clínica), mas também modelos complementares: erro total
analítico, período de janela seronegativo e modelo delta para uma estimativa inicial do impacto de resultados incertos no
orçamento de um serviço de sangue. Na perspetiva da incerteza diagnóstica reúnem‐se os modelos de exatidão diagnóstica
para a determinação da sensibilidade clínica, especificidade clínica e área sob a curva da característica de operação do recetor.
Complementariamente, são discutidos modelos para a estimação da concordância de resultados entre testes.
Para ilustrar os vários métodos de cálculo foram usados resultados de um imunoensaio quimiluminescente comum para
antivírus da hepatite C. Os cálculos mostram incerteza expandida em torno do “cutoff” de 21 a 36 %, tendo sido usado para a
estimativa da zona de rejeição [0.70, +[, na qual se registaram somente 0.19 % dos resultados de 9805 amostras. O erro total
analítico foi de 23 %. O modelo de período de janela estimou um período seronegativo de 97 dias, considerando a zona de
rejeição do imunoensaio. O valor delta indicou que a amostragem de doentes testada no “teste 1” tem uma menor
probabilidade para gerar resultados indeterminados, não tendo sido evidenciada uma diferença estatisticamente significativa
do “teste 1 e “teste 2” gerarem resultados indeterminados na amostra de não doentes. Os intervalos de confiança a 95 % para
a sensibilidade e especificidade clínicas foram, respetivamente, de ICS,95% = [88.3, 100 %] e ICE,95% = [98.5, 100 %], tendo‐se obtido
uma área sob a curva da característica de operação do recetor entre 0.99 e 1.00. A concordância total entre os resultados de
dois imunoensaios para antivírus da hepatite C foi de 98 a 99 %, tendo sido a concordância entre resultados positivos de 87.9 a
100 % e a concordância entre resultados negativos de 97.8 a 99.9 %. A aplicação do intervalo de confiança a 95 % às estimativas
de probabilidades e concordâncias é similar ao conceito de incerteza expandida da incerteza da medição.
Todos os métodos apresentados têm aplicação na avaliação da incerteza de medição e incerteza diagnóstica de um
imunoensaio, embora tenham condições de utilização diferentes. Recomenda‐se um esquema para a seleção dos modelos para
a estimativa da incerteza em laboratórios de rastreio de serviços de sangue, atendendo ao papel de cada modelo na segurança
pós‐transfusão.
Post‐transfusion safety of blood receptors is considered a major issue in the context of blood establishments. Considering the safety of blood receptors, these establishments must test a set of transmissible agents with a significant prevalence in the population. The role of screening tests’ results in the validation of the blood donations is critical for the assurance of posttransfusion safety. In the European Union countries, this screening is required by the Directive 2001/83/EC and in US by the “standards for blood banks and transfusion services”. Due to the uncertainty associated to screening tests’ results, a reported result can be false. A false positive result has no consequence in terms of the blood receptor safety. However, a false negative result has a major impact since there is a high probability of resulting in the infection of the blood component receptor. Consequently, screening laboratories should consider the evaluation of measurement uncertainty in the quality control schemes. However, the European Union or the US requests are generic when dealing with uncertainty. This thesis collects a set of articles discussing methods for the evaluation of measurement uncertainty in screening immunoassays such as complementary approaches and models of diagnostic accuracy. The Guide to the Expression of Uncertainty in Measurement (GUM) is not systematically used in blood establishments’ screening laboratories such as in other medical laboratories. However, other uncertainty estimations, namely based on statistical methods, have been used in blood establishments. Accordingly, the discussion includes not only models fulfilling GUM principles focused on uncertainty of ratio results close or equal to the “cutoff” (clinical decision value), but also complementary models: total analytical error, seronegative window period and the delta for a first estimation of the impact of uncertain results in the blood establishment budget. From the perspective of diagnostic uncertainty are reunited diagnostic accuracy models for determination of diagnostic sensitivity, diagnostic specificity and area under the receiver operating characteristic curve. Complementary are discussed models to agreement between tests’ results. To illustrate the evaluation of measurement uncertainty through the different methods, the results from a same anti‐hepatitis C virus chemiluminescence immunoassay are used. The computations reveal expanded uncertainties around “cutoff” from 21 to 36 % which were used to estimate the “rejection zone” [0.70, + [. Only 0.19 % of 9805 samples were in this region. The total analytical error was 23 %. Considering the immunoassay rejection zone, the window period model estimated a seronegative period of 97 days. The delta‐value indicated that the infected individuals’ sample tested in “test 1” is less likely to produce indeterminate results and has not been shown a statistically significant difference between “test 1” and “test 2” to generate indeterminate results in sample of healthy individuals. The diagnostic sensitivity and diagnostic specificity 95 % confidence intervals were [88.3, 100 %] and [98.5, 100 %], respectively, and the values of the area under the receiver operating characteristic curve obtained were between 0.99 and 1.00. The overall agreement of results between two anti‐hepatitis C virus immunoassays was from 98 to 99 %, the positive results agreement was from 87.9 to 100 % and the negative results agreement from 97.8 to 99.9 %. The 95 % confidence interval applied to the probabilities and agreements estimates is equivalent to the expanded uncertainty concept of measurement uncertainty. The presented methods have all application in the evaluation of measurement uncertainty and diagnostic uncertainty in an immunoassay, although under different conditions of use. A scheme for the selection of models for the estimation of uncertainty in screening laboratories in blood establishments is recommended, taking into account the role of each model in post‐transfusion safety.
Post‐transfusion safety of blood receptors is considered a major issue in the context of blood establishments. Considering the safety of blood receptors, these establishments must test a set of transmissible agents with a significant prevalence in the population. The role of screening tests’ results in the validation of the blood donations is critical for the assurance of posttransfusion safety. In the European Union countries, this screening is required by the Directive 2001/83/EC and in US by the “standards for blood banks and transfusion services”. Due to the uncertainty associated to screening tests’ results, a reported result can be false. A false positive result has no consequence in terms of the blood receptor safety. However, a false negative result has a major impact since there is a high probability of resulting in the infection of the blood component receptor. Consequently, screening laboratories should consider the evaluation of measurement uncertainty in the quality control schemes. However, the European Union or the US requests are generic when dealing with uncertainty. This thesis collects a set of articles discussing methods for the evaluation of measurement uncertainty in screening immunoassays such as complementary approaches and models of diagnostic accuracy. The Guide to the Expression of Uncertainty in Measurement (GUM) is not systematically used in blood establishments’ screening laboratories such as in other medical laboratories. However, other uncertainty estimations, namely based on statistical methods, have been used in blood establishments. Accordingly, the discussion includes not only models fulfilling GUM principles focused on uncertainty of ratio results close or equal to the “cutoff” (clinical decision value), but also complementary models: total analytical error, seronegative window period and the delta for a first estimation of the impact of uncertain results in the blood establishment budget. From the perspective of diagnostic uncertainty are reunited diagnostic accuracy models for determination of diagnostic sensitivity, diagnostic specificity and area under the receiver operating characteristic curve. Complementary are discussed models to agreement between tests’ results. To illustrate the evaluation of measurement uncertainty through the different methods, the results from a same anti‐hepatitis C virus chemiluminescence immunoassay are used. The computations reveal expanded uncertainties around “cutoff” from 21 to 36 % which were used to estimate the “rejection zone” [0.70, + [. Only 0.19 % of 9805 samples were in this region. The total analytical error was 23 %. Considering the immunoassay rejection zone, the window period model estimated a seronegative period of 97 days. The delta‐value indicated that the infected individuals’ sample tested in “test 1” is less likely to produce indeterminate results and has not been shown a statistically significant difference between “test 1” and “test 2” to generate indeterminate results in sample of healthy individuals. The diagnostic sensitivity and diagnostic specificity 95 % confidence intervals were [88.3, 100 %] and [98.5, 100 %], respectively, and the values of the area under the receiver operating characteristic curve obtained were between 0.99 and 1.00. The overall agreement of results between two anti‐hepatitis C virus immunoassays was from 98 to 99 %, the positive results agreement was from 87.9 to 100 % and the negative results agreement from 97.8 to 99.9 %. The 95 % confidence interval applied to the probabilities and agreements estimates is equivalent to the expanded uncertainty concept of measurement uncertainty. The presented methods have all application in the evaluation of measurement uncertainty and diagnostic uncertainty in an immunoassay, although under different conditions of use. A scheme for the selection of models for the estimation of uncertainty in screening laboratories in blood establishments is recommended, taking into account the role of each model in post‐transfusion safety.
Description
Keywords
Erro total analítico Exatidão diagnóstica GUM Imunoensaio de rastreio Incerteza da medição Incerteza diagnóstica Período de janela Segurança pós‐transfusão Serviço de sangue Valor delta Blood establishment Delta‐value Diagnostic accuracy Diagnostic uncertainty Measurement uncertainty Post‐transfusion safety Screening immunoassay Total analytical error Window period