i
CAMILA DE MELO CAMPOS
COMPARAÇÃO ENTRE QUATRO ÍNDICES DE MALIGNIDADE NA DISCRIMINAÇÃO PRÉ-
OPERATÓRIA DAS MASSAS ANEXIAIS
COMPARISION OF FOUR MALIGNANCY RISK INDICES IN THE PREOPERATIVE DISCRIMINATION
OF ADNEXAL MASSES
CAMPINAS 2014
ii
iii
UNIVERSIDADE ESTADUAL DE CAMPINAS
Faculdade de Ciências Médicas
CAMILA DE MELO CAMPOS
COMPARAÇÃO ENTRE QUATRO ÍNDICES DE MALIGNIDADE NA DISCRIMINAÇÃO PRÉ-
OPERATÓRIA DAS MASSAS ANEXIAIS
COMPARISION OF FOUR MALIGNANCY RISK INDICES IN THE PREOPERATIVE DISCRIMINATION
OF ADNEXAL MASSES
Dissertação apresentada à Pós-Graduação da Faculdade de Ciências Médicas da Universidade Estadual de Campinas para obtenção do Título de Mestra em Ciências da Saúde, área de concentração em Oncologia Ginecológica e Mamária.
Dissertation submitted to the Programme of Obstetrics and Gynecology of the Unicamp’s Health Sciences Faculty for obtaining the title of master in Health Sciences in the concentration area of Gynecologic and Breast Oncology
ORIENTADORA: PROFA. DRA. SOPHIE FRANÇOISE MAURICETTE DERCHAIN COORIENTADOR: PROF. DR. LUIS OTAVIO ZANATA SARIAN ESTE EXEMPLAR CORRESPONDE Á VERSÃO FINAL DA DISSERTAÇÃO DEFENDIDA PELA ALUNA CAMILA DE MELO CAMPOS E ORIENTADA PELA PROFA. DRA. SOPHIE FRANÇOISE MAURICETTE DERCHAIN
Assinatura do Orientador
CAMPINAS 2014
iv
Diagramação e Revisão: Assessoria Técnica do CAISM (ASTEC)
v
BANCA EXAMINADORA DA DEFESA
CAMILA DE MELO CAMPOS
ORIENTADORA: PROF. DRA. SOPHIE FRANÇOISE MAURICETTE DERCHAIN
COORIENTADOR: PROF. DR. LUIS OTÁVIO ZANATTA SARIAN
MEMBROS:
1.
2.
3.
Programa de Pós-Graduação em Tocoginecologia da Faculdade de Ciências Médicas da Universidade Estadual de Campinas
Data: 16 / 12 / 2014
vi
vii
RESUMO
A discriminação de tumores malignos entre mulheres com diagnóstico de massas
anexiais pode ser difícil devido a limitações na acurácia do exame
ultrassonográfico e à disponibilidade de pessoal especializado para realizá-lo. O
índice de risco de malignidade visa a simplificar e padronizar a rotina
ultrassonográfica para fornecer uma avaliação rápida e direta da massa anexial.
Neste estudo foi examinado o desempenho de quatro variações deste índice (IRM
1 a 4) em um centro terciário de assistência e pesquisa em câncer ginecológico
com a realização de exame ultrassonográfico por pessoal inserido em programa
de treinamento supervisionado. Método: 158 mulheres com diagnóstico de massa
anexial foram avaliadas antes da cirurgia utilizando-se as quatro variações do
IRM. O exame foi realizado por ultrassonografistas com níveis variados de
experiência e incluídos em programa de treinamento. Indicadores de desempenho
para os diferentes tipos de IRM foram calculados utilizando-se de metodologia
conhecida e o padrão-ouro para diagnóstico foi a análise anatomopatológica.
Resultados: A prevalência de tumores malignos foi de 32%. Pacientes com
tumores malignos eram mais idosas quando comparadas às pacientes com
diagnóstico de tumores benignos (idade média 45,9+15,0 anos versus 55,7+16,2;
p
viii
ovarianos era estágio I. Endometriomas foram as mais frequentes (11%) massas
anexiais não neoplásicas. Mulheres com tumores malignos apresentaram níveis
de CA125, escores de ultrassom e número de tumores com diâmetro >7 cm
significativamente maiores que mulheres com tumores benignos. Quando se
comparou o desempenho das variantes do IRM no melhor ponto de corte
determinado pela análise da curva ROC (receiver operator characteristic),
percebeu-se que as variantes do IRM apresentam desempenho semelhante na
população geral (pré e pós-menopausa). Entre as mulheres na pré-menopausa, a
melhor sensibilidade é obtida com o IRM2 (90%; 95% IC 83-97%) e com o IRM4
(89%; 95% IC 81-97%). A especificidade entre as diferentes variantes do IRM não
apresentou diferença significativa. O mesmo desempenho foi obtido entre as
variantes do IRM nas mulheres na pré e pós-menopausa. Foram também
analisados os indicadores de desempenho nas diferentes variantes do IRM nos
pontos de corte progressivos na população geral (pré e pós-menopausa). Os
pontos de corte recomendados pela literatura para os IRM1 a 3 é 200 e para o
IRM4 é 450. Nesses pontos de corte recomendados, a sensibilidade entre os
diferentes IRM variou entre 68% e 78% e a especificidade variou entre 82% e
87%. A pior correspondência entre valores do IRM e o resultado final
anatomopatologico foi obtido entre os tumores borderline, em que os tumores
foram classificados incorretamente em 50% dos casos utilizando o IRM1 e 3 e em
37% dos casos utilizando o IRM2 e 4. Proporções similares de tumores
classificados corretamente e incorretamente foram obtidos com as quatro
variantes do IRM. Os tumores epiteliais são mais bem classificados pelo IRM que
os não epiteliais. A taxa de falso negativo é maior entre os tumores do estroma:
ix
5/7 tumores de células da granulosa foram incorretamente classificados como
benignos entre as quatro variantes do IRM. Tumores borderlines foram
incorretamente classificados como benignos em 37% a 50% dos casos,
dependendo do IRM utilizado. Falsos negativos entre as quatro variantes do IRM
são maiores em mulheres com tumores de estágio 1 quando comparados com
mulheres em estágio mais avançado (p com valor significativo entre as quatro
variantes). Os IRM 1 e 3 classificaram incorretamente a maioria dos tumores
estágio 1 como benigno; IRM 2 classifica melhor tumores de estágio 1. É
importante ressaltar que 7 tumores de células da granulosa eram estágio 1.
Analisou-se a curva ROC para os diferentes IRM na discriminação das mulheres
entre tumores malignos e benignos. Os testes que compararam a área sobre a
curva de todas as curvas revelaram superioridade discreta do IRM4 sobre o IRM2
(p=0.06). Todos os outros testes realizados entre as curvas não obtiveram
resultado significativo. Conclusão: o IRM apresentou desempenho aceitável em
um centro terciário de assistência e pesquisa em câncer ginecológico, com
ultrassonografistas de conhecimento moderado e em treinamento. O equilíbrio
entre o desempenho e a viabilidade, devido à baixa complexidade da realização
do exame ultrassonográfico, favorece o IRM quando comparado a outros modelos
de triagem para avaliação de massas anexiais.
Palavras-chave: neoplasias ovarianas – diagnóstico; ultrassonografia.
x
xi
ABSTRACT
Discriminating women with ovarian malignancies among those with adnexal
masses may be difficult in medium resource settings due to limitations in
ultrasound accuracy and availability of specialized personnel. The Risk of
Malignancy Index (RMI) aims at simplifying and standardizing the ultrasound
routine in order to provide a fast and straightforward evaluation of the adnexal
mass. We examined the performance of four RMI variants (RMI 1 to 4) in a middle-
resources gynecologic cancer center, with ultrasound performed by personnel
under a training program. Methods: 158 referred due to an adnexal mass were
evaluated before surgery using the four RMI variants. Ultrasound was performed
by sonographers with variable expertise levels and enduring a training program.
Performance indicators for the RMI variants were calculated using standard
methodology and the gold standard was pathology of the adnexal mass. Results:
The prevalence of malignant tumor was 32%. Patients with malignant tumors were
significantly more aged than their counterparts with benign adnexal masses (mean
age 45.9+15.0 years versus 55.7+16.2; p7cm in
diameter than women with benign masses. When comparing the performance of
the RMI variants using the optimal cutoff points as determined with ROC analyses,
we notivce than in the general population (pre and postmenopausal women), RMI
xii
variants yielded similar performance indicators. In the subset of premenopausal
women, the best sensitivity was obtained with RMI 2 (90%; 95%CI 83-97%) and
RMI4 (89%; 95%CI 81-97%). Specificity for the RMI variants did not differ
significantly. Similar performance was obtained for the RMI variants in pre and
post-menopausal women. We then analyzed the performance indicators of RMI
variants at progressive cutoff points in the general (pre- and postmenopausal)
population. The standard (literature recommended) cutoff points for RMI 1 to 3 is
200 and for RMI 4 is 450. At these recommend cutoff points, the sensitivity of the
different RMI1 vary from 68% to 78% and specificity vary from 82% to 87%. The
worst correspondence between RMI values and final pathology was obtained for
borderline tumors, which were incorrectly classified in 50% of the cases using RMI
1 and 3 and 37% of the cases using RMI 2 and 4. Similar proportions of correctly
and incorrectly classified benign and malignant tumors were obtained with the four
RMI variants. Clearly, RMI classified epithelial tumors much better than it did with
non-epithelial tumors. The false negative rate was higher for stromal tumors: 5/7
granulosa cell tumors were incorrectly classified as benign by the four RMI
variants. Borderline tumors were also incorrectly classified as benign in 37-50% of
the cases depending on the RMI variant used. False negatives of for the RMI
variants are higher in women with stage 1 tumors compared to women with more
advanced stages (significant p values for all variants). RMI 1 and 3 incorrectly
classified the majority of stage 1 tumors as benign; RMI 2 was the variant that best
classified stage 1 tumors. It is worth noting that all 7 granulosa cell tumors were
stage 1. We analyzed the receiver–operating characteristics curve analysis of RMI
variants for the discrimination of women with malignant tumors from those with
xiii
benign tumors. The pairwise permutation tests comparing the AUC for the curves
revealed marginally significant superiority of RMI4 over RMI2 (p=0.06). All other
pairwise comparisons between the curves returned nonsignificant results.
Conclusions: RMI performed acceptably in a medium-resource setting where
sonographers had moderate expertise and/or were under training. The tradeoff
between performance and feasibility, due to lower ultrasound complexity, favors
RMI over other adnexal mass ultrasound-based triaging models.
Key words: ovarian neoplasia - diagnosis, ultrasound.
xiv
xv
SUMÁRIO
RESUMO................................................................................................................ vii
ABSTRACT ............................................................................................................. xi
SUMÁRIO............................................................................................................... xv
DEDICATÓRIA ..................................................................................................... xvii
AGRADECIMENTOS ............................................................................................ xix
SIGLAS E ABREVIATURAS ............................................................................... xxiii
LISTA DE SÍMBOLOS .......................................................................................... xxv
1. INTRODUÇÃO GERAL ..................................................................................... 1
2. OBJETIVOS .................................................................................................... 12
2.1 Objetivo Geral ......................................................................................... 12
2.2 Objetivos Específicos ............................................................................. 12
3. METODOLOGIA ............................................................................................. 13
4. CONCLUSÃO GERAL .................................................................................... 44
5. REFERÊNCIAS .............................................................................................. 45
6. ANEXOS ......................................................................................................... 53
xvi
xvii
DEDICATÓRIA
Às mulheres que contribuem para o meu crescimento pessoal e que não me
deixaram desistir de prosseguir meu caminho: Adriana, Sophie e Isa.
Ao meu pai, que sonha um sonho muito maior que o meu. Seu exemplo de caráter
me ensina a enxergar a vida com certezas e objetivos.
Ao meu irmão, que me ajuda a ser uma pessoa melhor.
xviii
xix
AGRADECIMENTOS
À equipe de ultrassonografia do CAISM, que me auxiliou na realização dos
exames e que cedeu o espaço para a coleta dos dados de forma amigável e
desinteressada.
Às pacientes que permitiram a realização dos exames e a coleta do CA125 e que
contribuíram para a realização desta pesquisa sem qualquer interesse
pessoal ou financeiro.
Aos médicos contratados da equipe de ultrassonografia, Rodrigo Jales e Danielle
Luminoso, que me orientaram no diagnóstico ultrassonográfico das pacientes
e estiveram comigo durante toda a coleta dos dados.
Aos funcionários do ambulatório de Ovário do CAISM, que realizaram a coleta do
CA125 das pacientes de forma precisa.
Aos médicos que atenderam as pacientes no ambulatório de Ovário do CAISM de
forma criteriosa e correta.
À equipe de patologia do CAISM, que realizou a análise anatomopatológica de
forma criteriosa e precisa.
À minha orientadora, Sophie Derchain, que de forma irretocável me ensinou tudo o
que aprendi neste projeto, além de estar ao meu lado nos momentos difíceis.
Ao meu co-orientador, Luis Otávio Sarian, que realizou as análises estatísticas e
nos ajudou na realização do artigo, sempre com boa vontade e entusiasmo.
Aos meus amigos, que estiveram do meu lado durante esta etapa importante,
estimulando-me a prosseguir.
À minha família, que me auxiliou a manter-me firme e não desistir dos meus
sonhos.
A Deus, que me deu um dom e sempre esteve ao meu lado neste projeto e em
todos em minha vida.
xx
xxi
Este estudo foi financiado por:
Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) processo
número 2012/15059-8.
xxii
xxiii
SIGLAS E ABREVIATURAS
AUC – Área Under the Curve
CAISM – Centro de Atenção Integral à Saúde da Mulher
CA125 – Cancer Antigen 125
cm – Centímetro(s)
CI – Confidence Interval
GE – General Electric
GI-RADS – Gynecologic Imaging Report and Data System
IBGE – Instituto Brasileiro de Geografia e Estatística
IOTA – International Ovarian Tumour Analysis
IRM/RMI – Índice de Risco de Malignidade / Risk of Malignancy
Index
LR/ LR2 – Logistic Regression / Logistic Regression 2
M – Menopausa
mm – Milímetro (s)
NPV – Negative Predictive Value
PPV – Positive Predictive Value
ROC – Receiver Operating Characteristic
SA – Subjective assessment
S – Size
SD – Standard Deviation
SR – Simple rules
Unicamp – Universidade Estadual de Campinas
U/ml – Unidade(s)/mililitro(s)
https://www.google.com.br/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=0CB4QFjAA&url=http%3A%2F%2Fwww.ibge.gov.br%2F&ei=N_xcVMmBIoqfNrW8gMAG&usg=AFQjCNGuNQz2gH-A3LkAod1HaFm3Qo6FTQ&sig2=Y4XiWBu9yZ81KbihsW5x9w&bvm=bv.79184187,d.eXY
xxiv
US – Ultrassom
xxv
LISTA DE SÍMBOLOS
% – Porcentagem
xxvi
1
1. INTRODUÇÃO GERAL
O câncer de ovário corresponde a 3,6% dos cânceres em geral [1]. Embora
não seja muito frequente, apresenta uma alta proporção de mortes por caso
detectado: em 2012, foram detectados 238.719 novos casos em todo o mundo,
dos quais 151.905 resultaram em mortes [1]. Foram estimados cerca de 5.680
casos novos de câncer de ovário no Brasil em 2014, com um risco estimado de 6
casos a cada 100 mil mulheres. Sem considerar os tumores da pele não
melanoma, o câncer de ovário é o oitavo mais incidente na maioria das regiões
brasileiras. Foram registradas 3.129 mortes pela doença em 2012 no Brasil [2]. A
incidência e mortalidade por câncer de ovário se mantiveram estáveis nas últimas
décadas. Observa-se um pequeno aumento da sobrevida em mulheres com
câncer, provavelmente relacionado à quimioterapia [3]. Devido à alta mortalidade
associada a essa doença, três grandes áreas devem ser priorizadas: esclarecer as
mulheres do risco para câncer de ovário, detectar a doença em estádios mais
iniciais e melhorar a qualidade dos tratamentos [4].
Programas de rastreamento para diferentes tumores, como mama ou colo
uterino, têm impacto significativo na mortalidade e detecção precoce. Porém,
estudos realizados em diferentes países não mostraram impacto significativo na
mortalidade na utilização do rastreamento para câncer de ovário, aumentando
risco de cirurgias desnecessárias e de suas complicações [5].
2
Para detectar o câncer de ovário em estádios iniciais, o exame mais
realizado é o ultrassom. Estima-se que 2,7% a 8% das mulheres apresentarão
cistos ovarianos ou massas anexiais durante a vida. Apesar de não haver dados
nacionais, acredita-se em uma incidência semelhante no Brasil. Assim, tumores
ovarianos são uma entidade comum que afeta mulheres em todas as idades. A
priori, a maioria dos tumores é benigna; na pré-menopausa muitas massas
anexiais são diagnosticadas como cistos funcionais ou neoplasias benignas.
Entretanto, 20% dos cânceres de ovário são detectados na menacme. Embora a
proporção de cânceres de ovário aumente na pós-menopausa [6], cistos
funcionais e tumores benignos ainda correspondem à maioria dos casos [7-10].
Menon et al. (2009) [11], em estudo em que 98.308 mulheres na pós-menopausa
foram randomizadas para realização de rastreamento com coleta de CA125 e
realização de ultrassom transvaginal, observaram que, das 942 mulheres
operadas por tumor anexial, 772 delas tiveram diagnóstico de neoplasia benigna
ovariana, o que corresponde a 80% dos tumores. Massas anexiais benignas
podem ser acompanhadas conservadoramente ou com realização de cirurgias
minimamente invasivas, como a laparoscopia, de menor hospitalização e
reabilitação mais precoce [8-13]. Por outro lado, o diagnóstico correto do câncer
de ovário é importante para garantir acesso a tratamentos adequados, uma vez
que a cirurgia inicial realizada interfere na sobrevida [14]. Nos casos de câncer de
ovário confirmados, o estadiamento cirúrgico é fundamental: avaliação cuidadosa
de todas as superfícies peritoneais, coleta de lavados peritoneais ou de ascite,
omentectomia infracólica, linfadenectomia das cadeias pélvicas e paraaórtica,
biópsia ou ressecção de quaisquer massas, lesão ou aderência suspeita, biópsias
3
aleatórias das superfícies peritoneais, histerectomia total, salpingooforectomia
bilateral e apendicectomia nos tumores mucinosos [15].
Para caracterizar o tumor anexial em benigno ou maligno é preciso utilizar
critérios que podem ser baseados em imagens ou associados a marcadores
tumorais e dados clínicos. Os métodos de diagnóstico de imagem de câncer de
ovário mais utilizados são ultrassonografia associada à ressonância magnética e a
tomografia computadorizada [16]. Porém, o grande número de exames de imagem
a que são submetidas as mulheres com suspeita de câncer de ovário antes da
cirurgia acaba retardando o tratamento, sendo esses exames frequentemente
desnecessários na prática clínica diária. Assim, a ultrassonografia baseada em
aspectos morfológicos das massas anexiais tem uma acurácia suficiente na
diferenciação das neoplasias malignas na maioria dos casos, permanecendo a
ressonância para tumores anexiais indeterminados e a tomografia para
estadiamento dos tumores malignos em casos selecionados [17].
Há várias décadas, pesquisadores têm estudado a criação de escores para
a classificação das massas anexiais a partir de critérios ultrassonográficos.
Gramberg et al. (1990) [18] avaliaram diferentes achados ultrassonográficos como
uni e multilocularidade, com e sem áreas sólidas em seu interior, até tumor sólido.
Sassone et al. (1991) [19] avaliaram a estrutura da parede do cisto e sua
espessura, presença ou ausência de septos e a ecogenicidade alta ou baixa. De
Priest el al. (1993) [20] avaliaram critérios como volume tumoral, estrutura da
parede do cisto, presença de septos, projeção papilar ou área sólida no interior do
cisto. Lerner et al. (1994) [21], por sua vez, incluíram a avaliação da sombra
acústica do cisto na classificação das massas anexiais. Ainda em 1994, Prömpeler
4
et al. [22] utilizaram critérios do Doppler para melhor caracterização das massas
anexiais, avaliando índice de resistência, pulsatilidade e velocidade arterial.
Conforme os critérios morfológicos e de avaliação por Doppler foram se
aprofundando, a análise estatística foi sendo desenvolvida com fórmulas mais
complexas, sendo utilizada a análise multivariada de regressão logística por Tailor
et al. (1997) [23]. As variáveis incluídas nesse estudo foram: idade, diâmetro
máximo do tumor, volume tumoral, presença de unilocularidade, ou de projeção
papilar, diferença na ecogenicidade e critérios do Doppler.
Paralelamente, desde a década de 1980, a dosagem sérica de marcadores
tumorais tem sido utilizada na diferenciação dos tumores anexiais [24]. O mais
utilizado é o CA125, uma glicoproteína localizada na superfície de muitas células
ovarianas cancerígenas [25]. Quando usado sozinho na distinção entre tumores
malignos e benignos em mulheres na menacme, tem uma baixa acurácia [16], já
que muitos fatores como ovulação, menstruação, endometriose, gestação podem
elevar seu nível sérico em mulheres saudáveis. Em mulheres com massa anexial
na pós-menopausa possui maior especificidade na diferenciação de tumores
malignos e benignos. Entretanto, o CA125 é negativo em 50% das neoplasias
restritas ao ovário e é positivo em 1,6% das mulheres menopausadas saudáveis
[24].
Com esses estudos, têm sido propostos vários métodos combinados com
métodos de imagem e utilização de marcadores para avaliação do risco de câncer
de ovário. O índice de risco de malignidade (IRM), escore baseado em achados do
ultrassom transvaginal, níveis do CA125 e menopausa (definida como amenorreia
por mais de um ano ou idade superior a 50 anos em mulheres submetidas à
5
histerectomia), é utilizado há décadas na discriminação dos tumores anexiais em
muitos países [26, 27, 28, 29]. Para calcular o IRM, o ultrassom recebe um escore
baseado nos aspectos morfológicos sugestivos de malignidade (presença de lesão
multilocular cística, áreas sólidas, lesões bilaterais, ascite ou metástase intra-
abdominal): cada aspecto equivale a um ponto no escore; a menopausa recebe
um escore para pré-menopausa e pós-menopausa e o CA125 entra com seu valor
total em U/mL, utilizando-se diferentes fatores de correção. O IRM é calculado
multiplicando os três escores (US x CA125 X menopausa) (quadro 1). No IRM 4, o
parâmetro tamanho do tumor é acrescido à formula (escore 1 se o tumor tiver o
maior diâmetro menor que 7cm e escore 2 se o tumor tiver o maior diâmetro maior
ou igual a 7cm). Embora haja pequenas diferenças na forma de calcular, os 4 IRM
parecem ter um desempenho semelhante na diferenciação pré-operatória das
massas anexiais. Para os IRM 1 a 3, o melhor ponto de corte foi de 200 e para o
IRM 4, de 450 [29,30].
6
Quadro 1: Diferenciação entre os quatro índices de risco de malignidade (IRM) conforme
cada autor [26,27,28,29].
IRM 1 IRM 2 IRM 3 IRM 4
Jacobs et al., 1990 Tingulstad et al., 1996 Tingulstad et al., 1999 Yamamoto et al., 2009
US 0 = 0
US 1 = 1
US >2 = 3
US 0 e 1=1
US >2 = 4
US 0 e 1=1
US >2 = 3
US 0 e 1=1
US >2 = 4
Pré-menopausa= M = 1
Pós-menopausa= M = 3
Pré-menopausa= M = 1
Pós-menopausa= M = 4
Pré-menopausa= M = 1
Pós-menopausa= M = 3
Pré-menopausa= M = 1
Pós-menopausa= M = 4
CA125 U/ml valor
diretamente aplicado na
fórmula
CA125 U/ml valor
diretamente aplicado na
fórmula
CA125 U/ml valor
diretamente aplicado na
fórmula
CA125 U/ml valor
diretamente aplicado na
fórmula
Maior diâmetro do tumor
Até 7 cm = S = 1
>7 cm = S = 2
Geomini et al. (2009) [31] avaliaram a acurácia de diferentes modelos de
diferenciação das massas anexiais em uma revisão sistemática. Foram incluídos
109 estudos na análise final, com 83 diferentes modelos de predição de
malignidade, somando-se 21.750 massas anexiais, sendo 15.490 benignas, 5.826
malignas e 434 borderlines. Eles verificaram uma sensibilidade global de 78% e
uma especificidade de 87% para o IRM em um valor de corte de 200 e concluíram
que os IRM 1 e 2 foram os melhores preditores de malignidade na discriminação
7
dos tumores anexiais. Podem ser utilizados como escolha na prática diária pela
sua simplicidade combinada a uma boa acurácia.
Por outro lado, nos modelos descritos acima, a utilização da dosagem
sérica do CA125 é um fator preponderante para avaliação dos tumores anexiais.
Por isso, alguns autores têm tentado identificar critérios ultrassonográficos que
melhor discriminem o câncer de ovário [32]. Timmerman et al. (2000) [33] e
Timmerman et al. (2008) [34] apresentaram os resultados de um grande estudo, o
International Ovarian Tumor Analysis (IOTA), que se utiliza de aspectos
ultrassonográficos e a idade da paciente. Estabeleceram um novo paradigma,
segundo o qual mais de 80% dos tumores anexiais poderiam ser adequadamente
classificados em benignos ou malignos, baseando-se em dez regras simples. Eles
se utilizam de cinco critérios baseados na imagem do ultrassom transvaginal para
definir um tumor como maligno (tumor sólido irregular, ascite, pelo menos quatro
estruturas papilares, tumor sólido irregular multilocular com um diâmetro maior que
100mm e alto teor de cor no exame Doppler colorido) e cinco para definir como
benigno (cisto uniloculado, a presença de componentes sólidos para o qual a
maior componente sólido é menor que 7mm de diâmetro, sombras acústicas,
tumor liso multilocular e não haver fluxo de sangue detectável no exame Doppler).
A presença de uma ou mais características malignas classifica o tumor como
maligno. Da mesma forma, a presença de uma ou mais características benignas
classifica o tumor como benigno. Porém, em cerca de 20% dos tumores não se
consegue identificar apenas critérios malignos ou benignos, devendo-se nesses
casos recorrer à experiência do ultrassonografista que irá classificar os tumores
segundo uma avaliação subjetiva [33-35].
8
Amor et al. (2011) [36] realizaram um estudo prospectivo multicêntrico
incluindo 432 massas anexiais em 372 mulheres. O objetivo desse estudo foi
aplicar na prática clínica diária um sistema de classificação baseado em achados
ultrassonográficos para homogeneizar o léxico e facilitar a comunicação entre os
ultrassonografistas, o Gynecologic Imaging Report and Data System (GI-RADS).
Ele utilizou como parâmetros os achados ultrassonográficos preditivos de
malignidade (ascite, áreas sólidas, septos grossos e projeções papilares) para
classificar as massas anexiais desde as definitivamente benignas (GI-RADS 1 –
probabilidade de malignidade de 0%) até muito provavelmente malignas (GI-RADS
5 – probabilidade de malignidade > 20%). Essa classificação auxilia na referencia
das pacientes com massas anexiais, desde o tratamento conservador para as
massas classificadas como GI-RADS 1 até o encaminhamento ao oncologista
ginecológico para as classificadas como GI-RADS 4 ou 5. Nesse estudo,
observou-se sensibilidade de 99,1% (95% IC, 95,1%–99,8%), especificidade de
85,9% (95% IC, 81,7%–89,3%) para a classificação das massas anexiais com alto
risco de malignidade, apresentando bom desempenho e podendo ser utilizado na
prática clínica diária. A crítica em relação a esse modelo de classificação é em se
basear também no léxico do IOTA e da complementação diagnóstica por
ultrassonografistas experientes, que utilizaram-se da avaliação subjetiva e de
padrão de reconhecimento para elucidação diagnóstica.
Em 2012 analisamos os critérios do IOTA em mulheres brasileiras com
massa anexiais. O estudo foi realizado com 103 mulheres portadoras de 110
tumores anexiais, sendo 31 malignos e 79 benignos. Dentre esses casos, os
critérios estabelecidos por Timmerman et al. (2010) [35] foram aplicáveis a
9
91(82%) tumores, com uma especificidade de 87% e uma sensibilidade de 90%.
Entretanto, 19 (18%) não foram classificáveis pelas regras simples [37]. Na prática
clínica diária nem sempre está disponível um ultrassonografista experiente, que foi
definido por Timmerman et al. (1999) [38] como o profissional com pelo menos
5.000 exames de ovário realizados no período de oito anos. Os autores do IOTA
procuraram então identificar outros modelos que não as regras simples para
serem utilizados por profissionais menos experientes. Após a validação interna de
11 modelos matemáticos, os pesquisadores concluíram que todos apresentam
resultados similares para discriminação das massas anexiais [39]. Entre os
diferentes modelos matemáticos, os modelos de regressão logística (LR) 1 e 2
foram amplamente utilizados e validados. Estão incluídos na avaliação critérios
objetivos como idade da paciente (em anos), presença de ascite, presença de
fluxo de sangue no interior da projeção papilar, máximo diâmetro do componente
sólido (em milímetros, até 50 mm), irregularidade no interior da parede do cisto, a
presença de sombra acústica, história pessoal de câncer de ovário, uso atual de
terapia hormonal, maior diâmetro da lesão em mm, presença ou ausência de dor
ao exame, presença de tumoração sólida e o escore de índice de cor (de 1 a 4).
Comparado o desempenho dos dois modelos (o LR1 com doze variáveis e o LR2
contendo as seis primeiras variáveis) observou-se que ambos apresentam
resultados similares, sendo o LR2 mais facilmente utilizável [39-41]. Com um valor
de LR2 > 10%, os tumores são classificados como de alto risco para doença
maligna [42,43]. Entretanto, a utilização desses modelos exige um conhecimento
adequado do léxico do IOTA.
10
Vários estudos sugerem que o IRM é um método mais facilmente utilizável
que o LR2, já que o IRM pode ser calculado sem o auxílio de um computador e
com achados morfológicos ultrassonográficos simples e validados, com alta
acurácia [44, 45]. Recentemente, Aktürk et al (2011) [30] avaliaram a performance
dos diferentes IRM (1,2,3 e 4) em 100 mulheres operadas por tumores anexais, e
chegaram à conclusão que todos os índices podem ser utilizados como preditor de
malignidade, sendo um método simples em sua realização e de alta acurácia. Van
der Akker et al. (2011) [44] validaram o IRM4 em comparação ao IRM3 na
discriminação das massas anexiais em estudo com 643 pacientes apresentando
469 tumores benignos, 101 tumores malignos e 73 tumores borderlines;
concluíram que o IRM3 apresenta melhor acurácia quando comparado ao IRM4. O
IRM3 teve sensibilidade de 76%, especificidade de 82% e acurácia de 81%
enquanto o IRM4 apresentou sensibilidade de 74%, especificidade de 79% e
acurácia de 78%. Ainda em 2011, Hakansson et al. [46] apresentaram um estudo
avaliando a performance do IRM3 na discriminação de 778 mulheres
diagnosticadas com tumor anexial. Em um ponto de corte de 200, o RMI3
apresentou sensibilidade de 92% e especificidade de 82% e valor preditivo
positivo e negativo de 62% e 97%, respectivamente. Em 2014, Abdulrahman et al.
[45] avaliaram o desempenho dos IRM 1, 2 e 3 na discriminação das massas
anexiais em 247 mulheres com diagnóstico de massa anexial e concluíram que os
RMI1 e 2 foram melhores preditores de malignidade que o IRM3 (utilizando-se o
ponto de corte de 200, o IRM1 apresentou sensibilidade de 66% e especificidade
de 91%, o IRM2 apresentou teve sensibilidade de 74% e especificidade de 79%
enquanto o IRM3 apresentou sensibilidade de 68% e especificidade de 85%).
11
Todos esses estudos concluíram que o IRM é um método simples de ser utilizado,
amplamente validado e de boa acurácia para discriminação das massas anexiais.
Vários estudos validaram o IRM em muitos países com bons resultados;
porém, não foram encontrams na literatura (Pubmed e Scielo) estudos que
avaliassem a acurácia dos IRMs no Brasil. Em serviços nacionais de atenção
primária é possível realizar ultrassonografia e dosagem sérica do marcador CA125
em mulheres com massas anexiais. Na atenção secundária, mulheres com
tumores anexiais benignos podem ser adequadamente tratadas, enquanto
mulheres com câncer de ovário se beneficiariam com encaminhamento e
tratamento em unidades de atenção terciária. Assim, comparar os diferentes IRMs
em mulheres brasileiras poderá trazer benefícios importantes na validação desses
métodos para a estruturação da atenção à saúde e melhoria no diagnóstico em
mulheres com massas anexiais.
12
2. OBJETIVOS
2.1 Objetivo Geral
Comparar o desempenho dos diferentes índices de risco de malignidade
(IRM) em mulheres na pré e pós-menopausa com massas anexiais submetidas à
cirurgia.
2.2 Objetivos Específicos
Avaliar a distribuição das mulheres com massa anexial segundo o
diagnóstico histológico, a idade, estado menopausal, antecedente familiar de
câncer de mama e ovário, a concentração sérica de CA125, o escore de
ultrassom e o tamanho do tumor.
Avaliar o desempenho dos diferentes IRM em mulheres com massas
anexiais na pré e na pós-menopausa segundo o ponto de corte definido pela
curva ROC.
Avaliar o desempenho dos diferentes IRM nos pontos de corte
estabelecidos pela literatura.
Avaliar a proporção de falsos positivos e falsos negativos segundo o tipo
histológico e o estádio.
13
3. METODOLOGIA
Performance of Risk of Malignancy Index (RMI) at discriminating malignant tumors in
women with adnexal masses in an ultrasound training center.
Camila Campos, MD(a)
Luis Otávio Sarian, MD, PhD(b)
Rodrigo Jales, MD, PhD(c)
Caio Hartman, MD(a)
Karla Araujo, MD(a)
Denise Pitta, Biologist(d)
Adriana Yoshida, MD(a)
Liliana Andrade, MD, PhD(e)
Sophie Derchain, MD, PhD(b)
Post Graduating Program in Tocogynecology(a), Department of Obstetrics and Gynecology
of the Faculty of Medical Sciences (b), Section of Ultrasonography, Prof. Dr. Jose
Aristodemo Pinotti Women’s Hospital, CAISM(c), Special Procedures Laboratory, Prof. Dr.
Jose Aristodemo Pinotti Women’s Hospital, CAISM (d), Department of Pathology(e),
Faculty of Medical Sciences, State University of Campinas – Unicamp, Campinas, São
Paulo, Brazil.
Correspondence to: Dr S. Derchain, Department of Obstetrics and Gynecology, Faculty of
Medical Sciences, PO Box 6111 State University of Campinas – UNICAMP, Zip Code
13083-970, Campinas, SP, Brazil (e-mail: [email protected]).
Short running title: risk of malignancy index in Brazilian women with adnexal mass
Type of article: original research
mailto:[email protected]).mailto:[email protected]).
14
Carta de submissão
Manuscript 15-01068 Version 1 Performance of Risk of Malignancy Index (RMI) at
discriminating malignant tumors in women with adnexal masses in an ultrasound
training center.
Dear Prof Derchain and coauthors,
Your manuscript, "Performance of Risk of Malignancy Index (RMI) at discriminating
malignant tumors in women with adnexal masses in an ultrasound training center.,"
has been assigned manuscript number 15-01068 Version 1. Please include this
number in the subject line of all e-mail correspondence regarding this manuscript.
(Note that this manuscript number is different from the TMP# your paper was
assigned during the submission phase.)
Your manuscript will undergo a quality assurance check to determine if it meets
requirements to be sent for review. If there are any problems or questions the
corresponding author will be notified.
Upon receipt of the reviewers' comments, the corresponding author will be notified
with a decision. If the initial reviewers were not in agreement, the editor may
request an opinion from an additional reviewer or a statistical reviewer before a
final decision is made. Please allow at least 1 month for a complete review of your
manuscript.
To see the status of your manuscript, log on to:
http://www.rapidreview.com/AIUM2/author.html
If your manuscript is accepted, it will be published in both the printed and online
versions of the Journal of Ultrasound in Medicine.
Thank you for submitting your manuscript.
Sincerely,
Editorial Assistant
Journal of Ultrasound in Medicine
American Institute of Ultrasound in Medicine
14750 Sweitzer Lane, Suite 100
Laurel, Maryland 20707-5906
http://www.rapidreview.com/AIUM2/author.html
15
USA
E-mail: [email protected]
mailto:[email protected]
16
Abstract
Objective: We examined the performance of four RMI variants (RMI 1 to 4) in a middle-
resources gynecologic cancer center, with ultrasound performed by personnel under a
training program. Methods: 158 women referred due to an adnexal mass were evaluated
before surgery using the four RMI variants. Ultrasound was performed by sonographers
with variable expertise levels and enduring a training program. We compared the
performance of the four RMI variants using receiver operator curve (ROC) analyses
followed by the calculation of sensitivity, specificity, positive and negative likelihood ratios
(LR+, LR-) using as gold standard the pathology of the adnexal mass. Results: Among the
158 women with adnexal masses included in this study, 51 (32%) had malignant tumors, 26
(51%) of them, stage I. All RMI variants performed similarly (accuracy ranging 74-83%),
regardless of menopausal status. Considering all women included, the LR+ of the four RMI
range from 3.52 to 4.41. In subset analyses, all RMI variants had decreased sensitivity for
stage 1 malignant tumors and for those with non-epithelial histology. Conclusions: The
four RMI performed acceptably in a medium-resource setting where sonographers had
moderate expertise and/or were under training. This is due to the good tradeoff between
performance and feasibility, since RMI ultrasound protocols are of low complexity.
Key words: ovarian tumor, malignancy, diagnostic, ultrasonography, CA125,
17
Introduction
There is no current strategy for ovarian cancer screening [1, 2], and it has been
demonstrated that women want some type of exam that could allow for early detection of
the disease [3]. Ultrasound is a widely available exam, and with it approximately 2.7% to
8% of women will be diagnosed with an adnexal mass at some point in life [4, 5, 6]. As a
result, one in ten women are still being operated for an adnexal mass in life, and devising
strategies for better selecting women who will derive a benefit from a surgical approach –
which must be relatively inexpensive and simple enough to promote widespread acceptance
by the medical community - is necessary [7, 8]. This is especially true in medium income
countries such as Brazil, where the demographics are now close to that of developed
countries but human and economic resources are still scarce. Approximately 5.680 ovarian
cancers are expected in Brazil in 2014, with an estimated risk of 6 cases/100,000 women
[9]. In excess of 3,000 deaths due to the disease were recorded in Brazil during 2012 [9].
In the last 30 years, several models including tumor makers and ultrasound (US)
descriptors and scores have been made in the field of better characterizing adnexal masses,
i.e. discriminating clinically relevant adnexal tumors from the vast majority of benign
masses. All these prediction models are currently undergoing testing as potential tools for
discerning the 20% to 35% of adnexal masses that are malignant ovarian tumors [10-15]
Since 1990, several mathematical models or scoring systems have been developed
to be used for discrimination between benign and malignant adnexal masses [12, 15-21].
Encouraging results were obtained with the Risk of Malignancy Index (RMI), which was
first developed in 1990 and received subsequent adjustments during the last twenty years
[11, 16-18, 22-24], and with a variety of models developed by the International Ovarian
18
Tumour Analysis’ (IOTA), notably the simple rules (SR), subjective assessment (SA) and
the logistic regression model (LR2) [15, 25]. IOTA studies suggested that SA, LR2, and SR
may perform better than RMI [15, 26, 27] in premenopausal women. In a previous study,
based in the IOTA results [28], we tested the SR in 103 women, and obtained a sensitivity
of 90%, specificity of 87%, positive predictive value (PPV) of 69% and negative predictive
value (NPV) of 97% [29]. However, 17.3% of the women had adnexal tumors not
classifiable by the SR, which prompted the need for an experienced sonographer, a
professional not widely available in our country. On the other hand, the RMI is a scoring
system that is derived from a formula that combines menopausal status with serum CA125
and US variables of low complexity [11, 16-18, 22, 30, 31]. Because US variables used in
RMI are much simpler than those used in IOTA models, and because RMI includes easily
obtainable laboratorial data (CA125 levels), it is sensible to infer that these models are
better suited for medium income settings.
In this study, we examine whether the outstanding results obtained and reported by
RMI creators are reproducible in a different set of pre- and postmenopausal Brazilian
women with adnexal masses and who underwent a surgical intervention due to these
masses. We also examined the factors associated with RMI failure at diagnosing malignant
tumors and at ruling out malignancy, such as tumor histological type and stage.
19
Subjects and methods
Patient selection
This is an analysis of prospectively collected data on 158 non-consecutive women
subjected to surgery due to an adnexal mass. Women had been referred to the gynecologic
oncology clinics of Campinas State University, Brazil, due to an adnexal mass detected
through sonography or clinical examination from January 2010 through January 2014.
At the first visit, women were informed that surgery had to be performed to treat her
adnexal mass. After the initial interview, including an explanation about the study’s
research methods and purpose, all women gave written informed consent to participate. An
ultrasound evaluation was scheduled and peripheral blood was collected for serum
measurements of the CA125 tumor marker. Patients underwent surgical intervention and
the pathologic specimens were sent for histopathological analysis. The study was approved
by the faculty’s research ethics committee under number 008/2010.
Ultrasound examination
Ultrasound evaluations were performed in the Ultrasound Technical Section of
UNICAMP, using one of the ultrasound machines available in the section: Accuvix V10
(Medison Corporation Ltd, Seoul, South Korea), Nemio XG (Toshiba Corporation, Tokyo,
Japan) and Voluson Expert 730 (GE Healthcare Ultrasound, Milwaukee, WI, USA), all
equipped with convex, endovaginal, broadband and high-resolution multifrequency
transducers, and all with amplitude spectral Doppler capability. The evaluation was
performed by physicians with variable expertise levels at assessing adnexal masses. For the
present study, the same physician performed and evaluated the ultrasound for each case.
The US scores were evaluated by that physician prospectively. All performing
20
sonographers were in a training program in gynecologic sonography for at least two years,
and all exams were performed under the supervision of a senior staff member, with a
minimum expertise of 5.000 exams. Ultrasound evaluation was performed with the woman
in a supine position. Initially we used a trans-abdominal approach, with the woman’s
bladder full; she was then asked to empty her bladder, and we performed a supplementary
transvaginal examination. Adnexal masses were described according to origin
(ovarian/extraovarian); position (right/left/bilateral); number of lesions; type of lesions
(unilocular/unilocularsolid/ multilocular/multilocular-solid), size in three dimensions
(longitudinal, anteroposterior and transverse diameters); volume (calculated electronically
by the ultrasound device which multiplicates the longitudinal, anteroposterior and
transversal diameters by the constant 0.52); presence and size of the largest solid
component (three diameters); presence and measurement of fluid volume in the posterior
cul-de-sac; and presence and location of lesions suggestive of metastases. Patients
presenting with at least one adnexal mass were eligible for inclusion in the study and when
there are more than one mass, the mass with the most complex morphology or, in cases of
similar morphology, the largest one, was considered, for statistical analyses as suggested by
Sayasneh et al. (2013) [32]. More than one adnexal mass was detected in 20 women.
21
Risk of Malignancy Index variants
The RMI is a scoring system that is derived from a formula that combines
menopausal status with serum CA125 and ultrasound variables. An ultrasound (US) score
is assigned for the following features suggestive of malignancy: the presence of
multilocular cystic lesion, solid areas, bilateral lesion, ascites, and intra-abdominal
metastasis. The presence of each of the previous parameters adds one point to the US score.
Based on the data obtained, four variants of the RMI (RMI 1, 2, 3 and 4) were calculated
for pre and post-menopausal women according to the original criteria and following
Yamamoto et al. (2009) [22] and Akturk et al (2011) [23]. In brief, all RMI variants are
based on the multiplication of a ultrasound score (U; see details below) by an arbitrary
value given to menopausal status (M; see details below) by the CA125 levels. For RMI 4,
tumor size is added. The following parameters were used for the calculations of each RMI
variant: RMI 1 (Jacobs et al. 1990) [16]= U × M × CA125 (ultrasound score: 0 made U=0;
a score of 1 made U=1; a score of ≥2 made U=3); premenopausal status made M=1 and
postmenopausal M=3. RMI 2 (Tingulstad et al. 1996) [17]= U × M × CA125, where a total
ultrasound score of 0 or 1 made U=1, and a score of ≥2 made U=4; premenopausal status
made M=1 and postmenopausal M=4. RMI 3 (Tingulstad et al. 1999) [18]= U × M ×
CA125, where a total ultrasound score of 0 or 1 made U=1, and a score of ≥2 made U=3;
premenopausal status made M=1 and postmenopausal M=3. RMI 4 (Yamamoto et al. 2009,
Akturk et al., 2011) [22, 23] = U × M × S × CA125, where a total ultrasound score of 0 or 1
made U=1, and a score of ≥2 made U=4. Premenopausal status made M=1 and
postmenopausal status made M=4. A tumor size (single greatest diameter) of
22
CA125 measurement
Roche Automated analysis of CA125 was performed by electrochemiluminescence
using the Cobas e411 test (Roche Diagnostics GmbH, Mannheim, Germany) according to
the manufacturer’s instructions and using their reagents and equipment. Values were
expressed in units per milliliter (U/mL). Post-menopausal status was defined as more than
one year of amenorrhea or age greater than 50 years in women who undergone
hysterectomy.
Surgery and pathology analysis
Surgery for diagnosis and/or treatment was performed at our institution, and the
techniques and surgical procedures were chosen and performed according to medical
indication. The mean time elapsed between ultrasound examination and surgery was 73
days, ranging from 24h or less for emergency procedures to a maximum of 119 days. The
gold standard was the histopathologic diagnosis of surgical specimens, all performed in the
Department of Pathologic Anatomy of the UNICAMP School of Medicine, following the
guidelines of the World Health Organization International Classification of Ovarian Tumors
(McCluggage, 2011) [33]. For statistical purposes, borderline tumors were classified as
malignant. Malignant ovarian tumors were staged according to the FIGO staging system
2013 [34].
Statistical analyses
All statistical calculations were performed using the R Environment [35] for data
analyses. 95% confidence levels were used throughout and a p-value of less than .05 was
considered significant. We first compared the proportion of the main clinical and
23
pathological features according to the pathological status (malignant versus benign) of their
tumors using chi-squares for categorical data and the Kruskal-Wallis test for continuous
data. Next, we calculated the performance of the RMI variants for the detection of
malignant tumors using standard Receiver Operating Characteristics Curves. We then
pairwise-compared the areas under the curves (AUC) for the RMI variations using the
Venkatraman´s Projection-Permutation test. Next, we calculated performance indicators
(sensitivity, specificity and positive and negative likelihood ratios (LR+, LR-, respectivelly)
) using the cutoff values determined by ROC analyses. Then, we recalculated the
performance indicators at recommended cutoff points (for RMI 1 to 3 = 200 and for RMI 4
= 450; Yamamoto et al., 2009, Akturk et al., 2011) [22, 23].
Results
Table 1 lists the key clinical and pathological features of the women. The
prevalence of malignant tumor was 32%. Patients with malignant tumors were significantly
more aged than their counterparts with benign adnexal masses (mean age 45.9+15.0 years
versus 55.7+16.2; p7cm in diameter than women with benign masses.
In Table 2 we compare the performance of the RMI variants using the optimal
cutoff points as determined with ROC analyses. In the general population (pre and
postmenopausal women), RMI variants yielded similar performance indicators. In the
subset of premenopausal women, the best sensitivity was obtained with RMI 2 (90%;
24
95%CI 83-97%) and RMI4 (89%; 95%CI 81-97%). Specificity for the RMI variants did not
differ significantly. Similar performance was obtained for the RMI variants in pre and post-
menopausal women. The four RMI had similar LR+ ranging from 2.92 to 5.68.
Table 3 shows the performance indicators of RMI variants at progressive cutoff
points in the general (pre- and postmenopausal) population. The standard (literature
recommended) cutoff points for RMI 1 to 3 is 200 and for RMI 4 is 450. At these
recommend cutoff points, the sensitivity of the different IRM 1vary from 68% to 78% and
specificity vary from 82% to 87%. In this recommended cut off point, the LR+ was 4.0 for
all RMI variants.
Table 4 shows how RMI variants classified benign, borderline and malignant
ovarian tumors at recommended cutoff points. Values above reference correspond to false
positives for benign tumors true positives for borderline and malignant tumors. The worst
correspondence between RMI values and final pathology was obtained for borderline
tumors, which were incorrectly classified in 50% of the cases using RMI 1 and 3 and 37%
of the cases using RMI 2 and 4. Similar proportions of correctly and incorrectly classified
benign and malignant tumors were obtained with the four RMI variants.
Table 5 shows how the RMI variants classified non-epithelial and epithelial
malignant tumors. Clearly, RMI classified epithelial tumors much better than it did with
non-epithelial tumors.
Table 6 shows diagnostic failures (false positives and negatives) of RMI variants at
recommended cutoff points, according to tumor histology. IRM1 and 3 and IRM 2 and 4
showed similar false positive and false negative results. The false negative rate was higher
for stromal tumors: 5/7 granulosa cell tumors were incorrectly classified as benign by the
25
four IRM variants. As shown in Table 4, borderline tumors were also incorrectly classified
as benign in 37-50% of the cases depending on the RMI variant used.
Table 7 shows that false negatives of for the RMI variants are higher in women with
stage 1 tumors compared to women with more advanced stages (significant p values for all
variants). RMI 1 and 3 incorrectly classified the majority of stage 1 tumors as benign; RMI
2 was the variant that best classified stage 1 tumors. It is worth noting that all 7 granulosa
cell tumors were stage 1.
In figure 1 we show the receiver–operating characteristics curve analysis of RMI
variants for the discrimination of women with malignant tumors from those with benign
tumors. All pairwise comparisons between the curves returned nonsignificant results.
Discussion
Our study confirms that RMI is a valuable tool in medium resource settings such as
the typical Brazilian healthcare system. In this sample of women with adnexal masses, all
RMI variants performed similarly (accuracy ranging 74-83%), regardless of menopausal
status. At the standard cutoff points, the sensitivity and specificity of all RMI variants were
very good, with LR+ in excess of 4.0 for all variants. It is important to notice, however, that
RMI variants had decreased sensitivity for stage 1 malignant tumors and in women with
non-epithelial tumors.
In the scarce resource environment where this study has been developed, highly
trained sonographers are scarce, although the epidemiology concerning adnexal tumors is
rapidly matching that of developed regions of the globe [36]. According to our data, all
RMI variants proved sufficiently sensitive and specific at diagnosing malignancy for both
pre and postmenopausal women.
26
In our study, the AUC observed for the RMI 1 to 4 was 0.85, albeit RMI 4 AUC
was slightly higher than that of RMI 2. Van den Akker et al. (2011) [37] compared the RMI
3 and RMI 4 and both proved to be capable of discriminating benign and malignant adnexal
lesions with similar performances, both with AUC of 0.86. In the same year, Akturk et al.
(2011) [23] repeated the performance RMI4, but found no significant differences between
the four different malignancy risk indices. It is worth noting that our ROC analyses showed
that optimal cutoff points for premenopausal women are substantially lower than those
preconized for the general population. At the standard cut off levels, our results closely
reproduced, in a population with a diverse epidemiologic background, those described by
Geomini et al. (2009) [11] in a systematic review evaluating the accuracy of risk scores,
when 200 was used as the cutoff level. In that analysis, the pooled estimates for sensitivity
was 78% and 87% for specificity.
Better triaging tools and protocols can assist the referral process of women with
adnexal masses to healthcare facilities with the necessary capabilities and guarantee
potential surgical failures\ and/or unnecessary overload of oncology centers with women
harboring benign conditions (Miller and Ueland, 2012) [38]. We detected only minimal
performance variability between the four RMI variants in this analysis on a relatively
homogeneous set of women with adnexal masses, who were treated at a single institution
and thus subject to similar treatment protocols. RMI 4 was slightly superior to RMI 2, but
only by a very non-significant small margin. These findings are in accordance with
Yamamoto at al. (2009) [22], who demonstrated that RMI4 was better than RMI1, RMI2
and RMI3, using a cutoff value of 450 for RMI 4 and 200 for the other variants. They
observed that the sensitivity, specificity, positive predictive value, negative predictive value
of RMI4 were respectively 86%, 91%, 63% and 97.5%. We obtained a sensitivity of 83%,
27
specificity of 81%, positive predictive value of 84% and 60% negative predictive values
using RMI4.
In our study, of the 51 malignant tumors, 31 were of epithelial origin, 8 were
borderline ovarian tumors and 8 were germ cell or stromal tumors. Meray at al (2010) [39]
demonstrated that RMI1 is not adequate for the detection of malignancy in a population
with high prevalence of borderline or non-epithelial tumors. In a population with 30% of
non-epithelial tumors, the sensitivity, specificity and positive and negative predictive values
were 60%, 88%, 57.1 and 89,9%, respectively. When these non-epithelial tumors are
excluded from the performance analyses, these indicators change to 76.9%, 88.7%, 52.6%
and 95.9%, respectively.
With standard cutoff points, sensitivity of all RMI variants may be severely
compromised in premenopausal women harboring stage 1 disease, stromal tumors or even
both. Van Gorp et al. (2012) [40] obtained 76% sensitivity and 92.4% specificity in the
general population, but sensitivity decreased to 64.1% in premenopausal women. Similar
findings were reported by authors using IOTA models, in a study that included 18
specialized centers in six different countries [15]: the sensitivity in the general population
was 67.1%(95%CI 61.4 to 72.4) and the specificity was 90.6% (95%CI 76.7 to 79.7);
however, in the subset of premenopausal women, the sensitivity decreased to 53% (95%CI
46 to 61).
Our study is flawed by a relatively small sample size and by not discriminating the
sonographers that performed the study exams according to their level of expertise. On the
other hand, this is a single institution trial, with a relatively high percentage of stage 1
malignant tumors. As mentioned above, this particular group of patients poses a challenge
to triaging methods, and our study corroborates that IRM may be not as good at diagnosing
28
early stage disease and non-epithelial ovarian tumors as originally thought. Our conclusions
would be weakened due to the small sample size of the non-epithelial tumors. For RMI
purposes, an ultrasound score is assigned considering the following features suggestive of
malignancy: the presence of multilocular cystic lesion, solid areas, bilateral lesion, ascites,
intra-abdominal metastasis. Sharma and colleagues investigated 48,053 asymptomatic
women who underwent ultrasound examination, 4,367 of whom (9.1% (95% CI, 8.8-9.3%))
had abnormal adnexal morphology. The strongest association between ovarian morphology
and epithelial ovarian cancer was the presence of ‘solid’ elements. The relative risk of
epithelial ovarian cancer within 3 years of the scan in women with solid elements compared
to unilocular or multilocular cysts was increased 11.5 (95% CI, 5.9–22.5)-fold [41].
The relative simplicity of the ultrasound parameters used to render RMI is a strong
advantage. Importantly, RMI includes CA125 levels in its formulae, and CA125
determination is a standardized, easily reproducible, and relatively cheap procedure
available even in low resource settings. These features obviate the need for highly
specialized sonographers; our study clearly confirms that RMI may yield acceptable
performance even when ultrasound is done by sonographers under training, which was the
case in our center. RMI still misses early stage and borderline tumors, as well as non-
epithelial neoplasms. In conclusion, discriminating women with ovarian malignancies
among those with adnexal masses may be difficult in medium resource settings due to
limitations in ultrasound accuracy and availability of specialized personnel. In our study,
we found that the four RMI performed acceptably in a medium-resource setting where
sonographers had moderate expertise and/or were under training. This is due to the good
tradeoff between performance and feasibility, since RMI ultrasound protocols are of low
complexity.
29
Acknowledgement: This study was partially financed by the Research Support
Foundation of the State of São Paulo – Fapesp: number 2012/15059-8. The authors also
thank the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) for
financial support.
30
References
1. Moyer VA; U.S. Preventive Services Task Force. Screening for ovarian cancer:
U.S. Preventive Services Task Force reaffirmation recommendation statement. Ann
Intern Med. 2012 Dec 18;157(12):900-4.
2. Menon U, Griffin M, Gentry-Maharaj A. Ovarian cancer screening--current status,
future directions. Gynecol Oncol. 2014 Feb; 132(2):490-5.
3. Pavlik EJ, van Nagell JR Jr. Ovarian cancer screening--what women want. Int J
Gynecol Cancer. 2012 May;22 Suppl 1:S21-3.
4. Kobayashi H, Yamada Y, Sado T, et al. A randomized study of screening for
ovarian cancer: a multicenter study in Japan. Int J Gynecol Cancer. 2008 May-
Jun;18(3):414-20
5. Menon U, Gentry-Maharaj A, Hallett R, et al. Sensitivity and specificity of
multimodal and ultrasound screening for ovarian cancer, and stage distribution of
detected cancers: results of the prevalence screen of the UK Collaborative Trial of
Ovarian Cancer Screening (UKCTOCS). Lancet Oncol. 2009 Apr; 10(4):327-40.
6. Buys SS, Partridge E, Black A, et al. PLCO Project Team. Effect of screening on
ovarian cancer mortality: the Prostate, Lung, Colorectal and Ovarian (PLCO)
Cancer Screening Randomized Controlled Trial. JAMA. 2011 Jun 8;305(22):2295-
303
7. Amonkar SD, Bertenshaw GP, Chen TH, et al. Development and preliminary
evaluation of a multivariate index assay for ovarian cancer. PLoS One.
2009;4(2):e4599
http://www.ncbi.nlm.nih.gov/pubmed/22964825http://www.ncbi.nlm.nih.gov/pubmed/22964825http://www.ncbi.nlm.nih.gov/pubmed/24316306http://www.ncbi.nlm.nih.gov/pubmed/24316306http://www.ncbi.nlm.nih.gov/pubmed/22543915http://www.ncbi.nlm.nih.gov/pubmed/17645503http://www.ncbi.nlm.nih.gov/pubmed/17645503http://www.ncbi.nlm.nih.gov/pubmed?term=Menon%20U%5BAuthor%5D&cauthor=true&cauthor_uid=19282241http://www.ncbi.nlm.nih.gov/pubmed?term=Gentry-Maharaj%20A%5BAuthor%5D&cauthor=true&cauthor_uid=19282241http://www.ncbi.nlm.nih.gov/pubmed?term=Hallett%20R%5BAuthor%5D&cauthor=true&cauthor_uid=19282241http://www.ncbi.nlm.nih.gov/pubmed/19282241http://www.ncbi.nlm.nih.gov/pubmed/21642681http://www.ncbi.nlm.nih.gov/pubmed/21642681http://www.ncbi.nlm.nih.gov/pubmed/21642681http://www.ncbi.nlm.nih.gov/pubmed/19240799http://www.ncbi.nlm.nih.gov/pubmed/19240799
31
8. French College of Gynecologists and Obstetricians. [Recommendations for clinical
practice: Presumed benign ovarian tumors]. J (Gynecol Obstet Biol Reprod (Paris).
2013 Dec;42(8):856-66.
9. INCA 2014. [assessed in August 15th, 2014]
http://www.inca.gov.br/estimativa/2014/
10. Huchon C, Bats AS, Bensaïd C, et al. [adnexal masses management: a prospective
multicentric observational study]. Gynecol Obstet Fertil. 2008 Nov;36(11):1084-90
11. Geomini P, Kruitwagen R, Bremer GL, Cnossen J, Mol BW. The accuracy of risk
scores in predicting ovarian malignancy: a systematic review. Obstet Gynecol. 2009
Feb;113(2 Pt 1):384-94.
12. Terzic M, Dotlic J, Brndusic N, et al. Histopathological diagnoses of adnexal
masses: which parameters are relevant in preoperative assessment? Ginekol
Pol. 2013 Aug;84(8):700-8.
13. Kaijser J, Sayasneh A, Van Hoorde K, et al. Presurgical diagnosis of adnexal
tumours using mathematical models and scoring system: a systematic review and
met-analysis. Hum Reprod Update. 2014a May-Jun;20(3):449-62.
14. Kaijser J, Van Belle V, Van Gorp T, et al. Prognostic Value of Serum HE4 Levels
and Risk of Ovarian Malignancy Algorithm Scores at the Time of Ovarian Cancer
Diagnosis. Int J Gynecol Cancer. 2014b Sep;24(7):1173-80.
15. Testa A, Kaijser J, Wynants L, et al. Strategies to diagnose ovarian cancer: new
evidence from phase 3 of the multicentre international IOTA study. Br J Cancer.
2014 Aug 12;111(4):680-8
http://www.ncbi.nlm.nih.gov/pubmed?term=French%20College%20of%20Gynecologists%20and%20Obstetricians%5BCorporate%20Author%5Dhttp://www.ncbi.nlm.nih.gov/pubmed/24210710http://www.inca.gov.br/estimativa/2014/http://www.ncbi.nlm.nih.gov/pubmed/19155910http://www.ncbi.nlm.nih.gov/pubmed/19155910http://www.ncbi.nlm.nih.gov/pubmed?term=Terzic%20M%5BAuthor%5D&cauthor=true&cauthor_uid=24191504http://www.ncbi.nlm.nih.gov/pubmed?term=Dotlic%20J%5BAuthor%5D&cauthor=true&cauthor_uid=24191504http://www.ncbi.nlm.nih.gov/pubmed?term=Brndusic%20N%5BAuthor%5D&cauthor=true&cauthor_uid=24191504http://www.ncbi.nlm.nih.gov/pubmed/24191504http://www.ncbi.nlm.nih.gov/pubmed/24191504http://www.ncbi.nlm.nih.gov/pubmed?term=Kaijser%20J%5BAuthor%5D&cauthor=true&cauthor_uid=24327552http://www.ncbi.nlm.nih.gov/pubmed?term=Sayasneh%20A%5BAuthor%5D&cauthor=true&cauthor_uid=24327552http://www.ncbi.nlm.nih.gov/pubmed?term=Van%20Hoorde%20K%5BAuthor%5D&cauthor=true&cauthor_uid=24327552http://www.ncbi.nlm.nih.gov/pubmed/?term=Presurgical+diagnosis+of+adnexal+tumours+using+mathematical+models+and+scoring+systems%3A+a+systematic+review+and+meta-analysishttp://www.ncbi.nlm.nih.gov/pubmed/24987915http://www.ncbi.nlm.nih.gov/pubmed/24987915http://www.ncbi.nlm.nih.gov/pubmed/24987915http://www.ncbi.nlm.nih.gov/pubmed/24937676http://www.ncbi.nlm.nih.gov/pubmed/24937676
32
16. Jacobs I, Oram D, Fairbanks J, Turner J, Frost C, Grudzinskas JG. A risk of
malignancy index incorporating CA 125, ultrasound and menopausal status for the
accurate preoperative diagnosis of ovarian cancer. Br J Obstet Gynaecol. 1990
Oct;97(10):922-9.
17. Tingulstad S, Hagen B, Skjeldestad FE, et al. Evaluation of a risk of malignancy
index based on serum CA125, ultrasound findings and menopausal status in the pre-
operative diagnosis of pelvic masses. Br J Obstet Gynaecol. 1996 Aug;103(8):826-
31.
18. Tingulstad S, Hagen B, Skjeldestad FE, Halvorsen T, Nustad K, Onsrud M. The
risk-of-malignancy index to evaluate potential ovarian cancers in local hospitals.
Obstet Gynecol. 1999 Mar;93(3):448-52.
19. Timmerman D, Van Calster B, Testa AC, et al. Ovarian cancer prediction in adnexal
masses using ultrasound-based logistic regression models: a temporal and external
validation study by the IOTA group. Ultrasound Obstet Gynecol. 2010a
Aug;36(2):226-34.
20. Lu KH, Patterson AP, Wang L, et al. Selection of potential markers for epithelial
ovarian cancer with gene expression arrays and recursive descent partition analysis.
Clin Cancer Res. 2004 May 15;10(10):3291-300.
21. Jokubkiene L, Sladkevicius P, Valentin L. Does three-dimensional power Doppler
ultrasound help in discrimination between benign and malignant ovarian masses?
Ultrasound Obstet Gynecol. 2007 Feb;29(2):215-25.
http://www.ncbi.nlm.nih.gov/pubmed?term=Jacobs%20I%5BAuthor%5D&cauthor=true&cauthor_uid=2223684http://www.ncbi.nlm.nih.gov/pubmed?term=Oram%20D%5BAuthor%5D&cauthor=true&cauthor_uid=2223684http://www.ncbi.nlm.nih.gov/pubmed?term=Fairbanks%20J%5BAuthor%5D&cauthor=true&cauthor_uid=2223684http://www.ncbi.nlm.nih.gov/pubmed?term=Turner%20J%5BAuthor%5D&cauthor=true&cauthor_uid=2223684http://www.ncbi.nlm.nih.gov/pubmed?term=Frost%20C%5BAuthor%5D&cauthor=true&cauthor_uid=2223684http://www.ncbi.nlm.nih.gov/pubmed?term=Grudzinskas%20JG%5BAuthor%5D&cauthor=true&cauthor_uid=2223684http://www.ncbi.nlm.nih.gov/pubmed/2223684http://www.ncbi.nlm.nih.gov/pubmed?term=Tingulstad%20S%5BAuthor%5D&cauthor=true&cauthor_uid=8760716http://www.ncbi.nlm.nih.gov/pubmed?term=Hagen%20B%5BAuthor%5D&cauthor=true&cauthor_uid=8760716http://www.ncbi.nlm.nih.gov/pubmed?term=Skjeldestad%20FE%5BAuthor%5D&cauthor=true&cauthor_uid=8760716http://www.ncbi.nlm.nih.gov/pubmed/8760716http://www.ncbi.nlm.nih.gov/pubmed?term=Tingulstad%20S%5BAuthor%5D&cauthor=true&cauthor_uid=10074998http://www.ncbi.nlm.nih.gov/pubmed?term=Hagen%20B%5BAuthor%5D&cauthor=true&cauthor_uid=10074998http://www.ncbi.nlm.nih.gov/pubmed?term=Skjeldestad%20FE%5BAuthor%5D&cauthor=true&cauthor_uid=10074998http://www.ncbi.nlm.nih.gov/pubmed?term=Halvorsen%20T%5BAuthor%5D&cauthor=true&cauthor_uid=10074998http://www.ncbi.nlm.nih.gov/pubmed?term=Nustad%20K%5BAuthor%5D&cauthor=true&cauthor_uid=10074998http://www.ncbi.nlm.nih.gov/pubmed?term=Onsrud%20M%5BAuthor%5D&cauthor=true&cauthor_uid=10074998http://www.ncbi.nlm.nih.gov/pubmed/10074998http://www.ncbi.nlm.nih.gov/pubmed?term=Timmerman%20D%5BAuthor%5D&cauthor=true&cauthor_uid=20455203http://www.ncbi.nlm.nih.gov/pubmed?term=Van%20Calster%20B%5BAuthor%5D&cauthor=true&cauthor_uid=20455203http://www.ncbi.nlm.nih.gov/pubmed?term=Testa%20AC%5BAuthor%5D&cauthor=true&cauthor_uid=20455203http://www.ncbi.nlm.nih.gov/pubmed/20455203http://www.ncbi.nlm.nih.gov/pubmed?term=Lu%20KH%5BAuthor%5D&cauthor=true&cauthor_uid=15161682http://www.ncbi.nlm.nih.gov/pubmed?term=Patterson%20AP%5BAuthor%5D&cauthor=true&cauthor_uid=15161682http://www.ncbi.nlm.nih.gov/pubmed?term=Wang%20L%5BAuthor%5D&cauthor=true&cauthor_uid=15161682http://www.ncbi.nlm.nih.gov/pubmed/15161682http://www.ncbi.nlm.nih.gov/pubmed/17201017http://www.ncbi.nlm.nih.gov/pubmed/17201017
33
22. Yamamoto Y, Yamada R, Oguri H, Maeda N, Fukaya T. Comparison of four
malignancy risk indices in the preoperative evaluation of patients with pelvic
masses. Eur J Obstet Gynecol Reprod Biol. 2009 Jun;144(2):163-7.
23. Aktürk E, Karaca RE, Alanbay I, et al. Comparison of four malignancy risk indices
in the detection of malignant ovarian masses. J Gynecol Oncol. 2011
Sep;22(3):177-82
24. Bouzari Z, Yazdani S, Ahmadi MH, et al. Comparison of three malignancy risk
indices and CA-125 in the preoperative evaluation of patients with pelvic masses.
BMC Res Notes. 2011 Jun 20;4:206.
25. Van Calster B, Timmerman D, Valentin L, et al. Triaging women with ovarian
masses for surgery: observational diagnostic study to compare RCOG guidelines
with an International Ovarian Tumour Analysis (IOTA) group protocol. BJOG.
2012 May;119(6):662-71.
26. Timmerman D, Schwärzler P, Collins WP, et al. Subjective assessment of adnexal
masses with the use of ultrasonography: an analysis of interobserver variability and
experience. Ultrasound Obstet Gynecol. 1999 Jan;13(1):11-6.
27. Timmerman D, Testa AC, Bourne T, et al. International Ovarian Tumor Analysis
Group. Logistic regression model to distinguish between the benign and malignant
adnexal mass before surgery: a multicenter study by the International Ovarian
Tumor Analysis Group. J Clin Oncol. 2005 Dec 1;23(34):8794-801.
28. Timmerman D, Ameye L, Fischerova D, et al. Simple ultrasound rules to
distinguish between benign and malignant adnexal masses before surgery:
prospective validation by IOTA group. BMJ. 2010b Dec 14;341:c6839.
http://www.ncbi.nlm.nih.gov/pubmed/19327881http://www.ncbi.nlm.nih.gov/pubmed/19327881http://www.ncbi.nlm.nih.gov/pubmed/19327881http://www.ncbi.nlm.nih.gov/pubmed/21998760http://www.ncbi.nlm.nih.gov/pubmed/21998760http://www.ncbi.nlm.nih.gov/pubmed?term=Bouzari%20Z%5BAuthor%5D&cauthor=true&cauthor_uid=21689405http://www.ncbi.nlm.nih.gov/pubmed?term=Yazdani%20S%5BAuthor%5D&cauthor=true&cauthor_uid=21689405http://www.ncbi.nlm.nih.gov/pubmed?term=Ahmadi%20MH%5BAuthor%5D&cauthor=true&cauthor_uid=21689405http://www.ncbi.nlm.nih.gov/pubmed/21689405http://www.ncbi.nlm.nih.gov/pubmed?term=Van%20Calster%20B%5BAuthor%5D&cauthor=true&cauthor_uid=22390753http://www.ncbi.nlm.nih.gov/pubmed?term=Timmerman%20D%5BAuthor%5D&cauthor=true&cauthor_uid=22390753http://www.ncbi.nlm.nih.gov/pubmed?term=Valentin%20L%5BAuthor%5D&cauthor=true&cauthor_uid=22390753http://www.ncbi.nlm.nih.gov/pubmed/22390753http://www.ncbi.nlm.nih.gov/pubmed?term=Timmerman%20D%5BAuthor%5D&cauthor=true&cauthor_uid=10201081http://www.ncbi.nlm.nih.gov/pubmed?term=Schw%C3%A4rzler%20P%5BAuthor%5D&cauthor=true&cauthor_uid=10201081http://www.ncbi.nlm.nih.gov/pubmed?term=Collins%20WP%5BAuthor%5D&cauthor=true&cauthor_uid=10201081http://www.ncbi.nlm.nih.gov/pubmed/?term=Timmerman+D%2C+Schwarzler+P%2C+Collins+WPhttp://www.ncbi.nlm.nih.gov/pubmed?term=Timmerman%20D%5BAuthor%5D&cauthor=true&cauthor_uid=16314639http://www.ncbi.nlm.nih.gov/pubmed?term=Testa%20AC%5BAuthor%5D&cauthor=true&cauthor_uid=16314639http://www.ncbi.nlm.nih.gov/pubmed?term=Bourne%20T%5BAuthor%5D&cauthor=true&cauthor_uid=16314639http://www.ncbi.nlm.nih.gov/pubmed?term=International%20Ovarian%20Tumor%20Analysis%20Group%5BCorporate%20Author%5Dhttp://www.ncbi.nlm.nih.gov/pubmed?term=International%20Ovarian%20Tumor%20Analysis%20Group%5BCorporate%20Author%5Dhttp://www.ncbi.nlm.nih.gov/pubmed/?term=Timmerman+D%2C+Testa+AC%2C+Bourne+T%2C+Ferrazzi+E%2C+Ameye+L%2C+Konstantinovic+ML%2C+Van+Calster+B%2C+Collins+WP%2C+Vergote+I%2C+Van+Huffel+S%2C+Valentin+L.http://www.ncbi.nlm.nih.gov/pubmed?term=Timmerman%20D%5BAuthor%5D&cauthor=true&cauthor_uid=21156740http://www.ncbi.nlm.nih.gov/pubmed?term=Ameye%20L%5BAuthor%5D&cauthor=true&cauthor_uid=21156740http://www.ncbi.nlm.nih.gov/pubmed?term=Fischerova%20D%5BAuthor%5D&cauthor=true&cauthor_uid=21156740http://www.ncbi.nlm.nih.gov/pubmed/21156740
34
29. Hartman CA, Juliato CR, Sarian LO, et al. Ultrasound criteria and CA 125 as
predictive variables of ovarian cancer in women with adnexal tumors. Ultrasound
Obstet Gynecol. 2012 Sep;40(3):360-6.
30. Håkansson F, Høgdall EV, Nedergaard L, et al. DANISH ‘PELVIC MASS’
OVARIAN CANCER STUDY. Risk of malignancy index used as a diagnostic tool
in a tertiary centre for patients with a pelvic mass. Acta Obstet Gynecol Scand. 2012
Apr;91(4):496-502.
31. Abdulrahman GO Jr, McKnight L, Lutchman Singh K. The risk of malignancy
index (RMI) in women with adnexal masses in Wales. Taiwan J Obstet
Gynecol. 2014 Sep;53(3):376-81.
32. Sayasneh A, Wynants L, Preisler J, et al. Multicentre external validation of IOTA
prediction models and RMI by operators with varied training. Br J Cancer. 2013 Jun
25;108(12):2448-54.
33. McCluggage WG. Morphological subtypes of ovarian carcinoma: a review with
emphasis on new developments and pathogenesis. Pathology. 2011;43(5):420-32.
34. Mutch DG, Prat J: 2014 FIGO staging for ovarian, fallopian tube and peritoneal
cancer. Gynecol Oncol 133 (3): 401-4, 2014.
35. Development Core Team R: R: A language and environment for statistical
computing. Vienna, Austria: R Foundation for Statistical Computing; 2011.
36. Globocan 2012. [Assessed September 29th, de 2014] http://globocan.iarc.fr.
37. Van den Akker PA, Zusterzeel PL, Aalders AL, et al. External validation of the
adapted Risk of Malignancy Index incorporating tumor size in the preoperative
evaluation of adnexal masses. Eur J Obstet Gynecol Reprod Biol. 2011
Dec;159(2):422-5.
http://www.ncbi.nlm.nih.gov/pubmed/?term=Hartman+CA%2C+Juliato+CR%2C+Sarian+LO%2C+Toledo+MC%2C+Jales+RM%2C+Morais+SShttp://www.ncbi.nlm.nih.gov/pubmed/?term=Hartman+CA%2C+Juliato+CR%2C+Sarian+LO%2C+Toledo+MC%2C+Jales+RM%2C+Morais+SShttp://www.ncbi.nlm.nih.gov/pubmed/22229703http://www.ncbi.nlm.nih.gov/pubmed/22229703http://www.ncbi.nlm.nih.gov/pubmed?term=Abdulrahman%20GO%20Jr%5BAuthor%5D&cauthor=true&cauthor_uid=25286794http://www.ncbi.nlm.nih.gov/pubmed?term=McKnight%20L%5BAuthor%5D&cauthor=true&cauthor_uid=25286794http://www.ncbi.nlm.nih.gov/pubmed?term=Lutchman%20Singh%20K%5BAuthor%5D&cauthor=true&cauthor_uid=25286794http://www.ncbi.nlm.nih.gov/pubmed/25286794http://www.ncbi.nlm.nih.gov/pubmed/25286794http://www.ncbi.nlm.nih.gov/pubmed?term=Sayasneh%20A%5BAuthor%5D&cauthor=true&cauthor_uid=23674083http://www.ncbi.nlm.nih.gov/pubmed?term=Wynants%20L%5BAuthor%5D&cauthor=true&cauthor_uid=23674083http://www.ncbi.nlm.nih.gov/pubmed?term=Preisler%20J%5BAuthor%5D&cauthor=true&cauthor_uid=23674083http://www.ncbi.nlm.nih.gov/pubmed/?term=multicentre+external+validation+sayasnehhttp://globocan.iarc.fr/http://www.ncbi.nlm.nih.gov/pubmed/21824712http://www.ncbi.nlm.nih.gov/pubmed/21824712http://www.ncbi.nlm.nih.gov/pubmed/21824712
35
38. Miller RW, Ueland FR. Risk of malignancy in sonographically confirmed ovarian
tumors. Clin Obstet Gynecol. 2012 Mar;55(1):52-64.
39. Meray O, Türkçüoğlu I, Meydanlı MM, Kafkaslı A. Risk of malignancy index is not
sensitive in detecting non-epithelial ovarian cancer and borderline ovarian tumor. J
Turk Ger Gynecol Assoc. 2010 Mar 1;11(1):22-6.
40. Van Gorp T, Veldman J, Van Calster B, et al. Subjective assessment by ultrasound
is superior to the risk of malignancy index (RMI) or the risk of ovarian malignancy
algorithm (ROMA) in discriminating benign from malignant adnexal masses. Eur J
Cancer. 2012 Jul;48(11):1649-56.
41. Sharma A, Apostolidou S, Burnell M, et al. Risk of epithelial ovarian cancer in
asymptomatic women with ultrasound-detected ovarian masses: a prospective
cohort study within the UK collaborative trial of ovarian cancer screening
(UKCTOCS). Ultrasound Obstet Gynecol. 2012 Sep;40(3):338-44.
http://www.ncbi.nlm.nih.gov/pubmed/22343229http://www.ncbi.nlm.nih.gov/pubmed/22343229http://www.ncbi.nlm.nih.gov/pubmed?term=Meray%20O%5BAuthor%5D&cauthor=true&cauthor_uid=24591890http://www.ncbi.nlm.nih.gov/pubmed?term=T%C3%BCrk%C3%A7%C3%BCo%C4%9Flu%20I%5BAuthor%5D&cauthor=true&cauthor_uid=24591890http://www.ncbi.nlm.nih.gov/pubmed?term=Meydanl%C4%B1%20MM%5BAuthor%5D&cauthor=true&cauthor_uid=24591890http://www.ncbi.nlm.nih.gov/pubmed?term=Kafkasl%C4%B1%20A%5BAuthor%5D&cauthor=true&cauthor_uid=24591890http://www.ncbi.nlm.nih.gov/pubmed/?term=Meray+and+2010+and+ovarian+masshttp://www.ncbi.nlm.nih.gov/pubmed/?term=Meray+and+2010+and+ovarian+masshttp://www.ncbi.nlm.nih.gov/pubmed?term=Van%20Gorp%20T%5BAuthor%5D&cauthor=true&cauthor_uid=22226481http://www.ncbi.nlm.nih.gov/pubmed?term=Veldman%20J%5BAuthor%5D&cauthor=true&cauthor_uid=22226481http://www.ncbi.nlm.nih.gov/pubmed?term=Van%20Calster%20B%5BAuthor%5D&cauthor=true&cauthor_uid=22226481http://www.ncbi.nlm.nih.gov/pubmed/?term=VAN+GORP+AND+VELDMAN+AND+CADRONhttp://www.ncbi.nlm.nih.gov/pubmed/?term=VAN+GORP+AND+VELDMAN+AND+CADRONhttp://www.ncbi.nlm.nih.gov/pubmed?term=Sharma%20A%5BAuthor%5D&cauthor=true&cauthor_uid=22911637http://www.ncbi.nlm.nih.gov/pubmed?term=Apostolidou%20S%5BAuthor%5D&cauthor=true&cauthor_uid=22911637http://www.ncbi.nlm.nih.gov/pubmed?term=Burnell%20M%5BAuthor%5D&cauthor=true&cauthor_uid=22911637http://www.ncbi.nlm.nih.gov/pubmed/22911637
36
Table 1. Key clinical features of women with ovarian benign and malignant tumors
Characteristic
Benign
N= 107
Malignant
N= 51
p-value
Age
Years, mean (SD)*
45.9 (15.0)
55.7 (16.2)
37
Table 2: Performance of RMI variants in pre- and posmenopausal women at cutoff points
determined by ROC analyses.
Group Index AUC Cut off Sensitivity(%) Specificity(%) Accuracy
(%)
LR+ LR-
All women RMI1 0.85 (0.78-0.91) 93.9 82 (75-90) 77 (67-88) 79 3.67 0.22
RMI2 0.85 (0.78-0.91) 195.7 78 (71–86) 82 (72-92) 81 4.41 0.26
RMI3 0.85 (0.78-0.91) 93.9 82 (75-90) 77 (65-86) 78 3.52 0.23
RMI4 0.85 (0.77-0.92) 250.4 83 (76-90) 81 (69-90) 81 4.29 0.21
Pre-menopause RMI1 0.84 (0.78-0.91) 93.9 70 (59-81) 88 (74-100) 83 5.68 0.34
RMI2 0.85 (0.74-0.96) 50.8 90 (83-97) 69 (59-84) 74 2.92 0.14
RMI3 0.84 (0.73-0.95) 93.9 70 (59-81) 89 (76-100) 76 3.46 0.32
RMI4 0.86 (0.72-0.98) 101.8 89 (81-97) 78 (63-93) 78 4.19 0.32
Post-
menopause
RMI1 0.81 (0.72-0.91) 238.5 74 (61-87) 78 (64-93) 77 3.46 0.32
RMI2 0.81 (0.71-0.91) 424.0 71 (57-85) 81 (67-95) 77 3.72 0.36
RMI3 0.81 (0.71-0.91) 238.5 74 (61-87) 79 (64-.93) 76 3.46 0.32
RMI4 0.79 (0.68-0.90) 848.0 73 (60-87) 82 (69-96) 78 4.19 0.32
AUC=area under the Receiver–operating characteristics curve , PPV = positive predictive
value, NPV=negative predictive value
38
Table 3: Performance comparison of RMI variants at progressing cutoff levels for the detection of malignant ovarian tumors
Cutoff Sensitivity Specificity LR+ LR-
RMI
1, 2,
3
RMI4 RMI1 RMI2 RMI3 RMI4 RMI1 RMI2 RMI3 RMI4 RMI1 RMI2 RMI3 RMI4 RMI1 RMI2 RMI3 RMI4
50 300 86 96 88 79 60 52 57 81 2.14 2.01 2.05 4.29 0.22 0.07 0.21 0.25
100 350 78 82 78 77 78 73 77 81 3.64 3.03 3.49 4.17 0.27 0.24 0.27 0.28
150 400 74 78 74 77 84 80 83 82 4.68 3.99 4.42 4.41 0.30 0.27 0.31 0.28
200* 450* 68 78 69 75 87 82 87 82 5.24 4.41 5.24 4.29 0.36 0.26 0.36 0.30
250 500 67 74 67 73 89 83 89 83 5.94 4.42 5.94 4.41 0.37 0.31 0.37 0.32
300 550 63 69 63 71 89 85 89 85 5.59 4.58 5.59 4.86 0.42 0.37 0.41 0.34
350 600 63 67 63 71 90 85 90 85 6.10 4.45 6.10 4.86 0.41 0.39 0.41 0.34
400 650 61 65 61 71 91 88 91 85 7.22 5.32 7.22 4.86 0.43 0.40 0.43 0.34
*Standard (literature recommended) cutoff points for pre- and postmenopausal women with adnexal masses
39
Table 4: Proportion of benign, borderline and malignant tumors at recommended cutoff
points for RMI variants’
RMI
variant
Stratum Total Pathological status
Benign (n=107) Borderline
(n=8)
Malignant (n=43)
RMI 1 < 200 109 93 (87%) 4 (50%) 12 (28%)
>200 49 14 (13%) 4 (50%) 31 (72%)
RMI 2 < 200 99 88 (82%) 3 (37%) 8 (19%)
>200 59 19 (18%) 5 (63%) 35 (81%)
RMI 3 < 200 109 93 (87%) 4 (50%) 12 (28%)
>200 49 14 (13%) 4 (50%) 31 (72%)
RMI 4 < 450 101 88 (82%) 3 (37%) 10 (23%)
>450 57 19 (18%) 5 (63%) 33 (77%)
RMI = Risk of Malignancy Index.
40
Table 5: Proportion of epithelial and non-epithelial ovarian malignant tumors at
recommended cutoff points for the RMI variants
Index Stratum Total Primary Ovarian Malignancy
NON EPITHELIAL
EPITHELIAL P
RMI 1 < 200 16 5 (63%) 10 (27%) 0.132
>200 32 3 (37%) 29 (73%)
RMI2 < 200 11 5 (63%) 6 (15%) 0.014
>200 37 3 (37%) 34 (85%)
RMI 3 < 200 16 5 (63%) 11 (27
Top Related