Arq Bras CardiolabcArquivos Brasileiros de CardiologiaArq. Bras. Cardiol.0066-782X1678-4170Sociedade Brasileira de Cardiologia - SBCPnDBGw5WhwwxD9qYhSnHZGFS0066-782X202500060031110.36660/abc.20240405 00311Artigo OriginalConcordância entre Framingham, Pooled Cohort Equations e Globorisk-LAC da estimação do risco cardiovascular no Brasil, 20130000-0003-4064-433XFontenelleLeonardo FerreiraConcepção e desenho da pesquisa e Análise e interpretação dos dadosAnálise estatística e Redação do manuscritoSartiThiago DiasConcepção e desenho da pesquisa e Análise e interpretação dos dadosRevisão crítica do manuscrito quanto ao conteúdo1QuinteGabriela CalloConcepção e desenho da pesquisa e Análise e interpretação dos dadosRevisão crítica do manuscrito quanto ao conteúdo2AlmeidaAna Paula Santana CoelhoConcepção e desenho da pesquisa e Análise e interpretação dos dadosRevisão crítica do manuscrito quanto ao conteúdo10000-0002-0987-368XMillJosé GeraldoConcepção e desenho da pesquisa e Análise e interpretação dos dadosRevisão crítica do manuscrito quanto ao conteúdo3Universidade Federal do Espírito SantoDepartamento de Medicina SocialVitoriaESBrasilUniversidade Federal do Espírito Santo (UFES) Departamento de Medicina Social, Vitoria, ES – BrasilUniversidade Federal do Espírito SantoPrograma de Pós-Graduação em Saúde ColetivaVitoriaESBrasilUniversidade Federal do Espírito Santo (UFES) Programa de Pós-Graduação em Saúde Coletiva, Vitoria, ES – BrasilUniversidade Federal do Espírito SantoPrograma de Pós-Graduação em Ciências FisiológicasVitoriaESBrasilUniversidade Federal do Espírito Santo (UFES) Programa de Pós-Graduação em Ciências Fisiológicas, Vitoria, ES – BrasilLeonardo Ferreira Fontenelle • Universidade Federal do Espírito Santo (UFES) – Departamento de Saúde Coletiva – Av. Marechal Campos, 1468. CEP 29047-105, Bonfim, Vitoria, ES – Brasil E-mail: leonardof@leonardof.med.br
Não há conflito com o presente artigo.
Marcio Bittencourt
28072025202512206e20240405150620242602202519032025This is an open-access article distributed under the terms of the Creative Commons Attribution LicenseResumoFundamento
Os escores de Framingham e Pooled Cohort Equations (PCE) nunca foram recalibrados para a população brasileira. O Globorisk-LAC foi derivado recentemente com uma metodologia análoga ao PCE, tendo sido recalibrado para os países da América Latina.
Objetivos
Descrever a concordância entre os escores de Framingham, PCE e Globorisk-LAC na estimação do risco cardiovascular em dez anos na população brasileira.
Métodos
Neste estudo de corte transversal, o risco foi estimado pelos três escores para pessoas com idade entre 40–74 anos sem doença cardiovascular, usando dados da Pesquisa Nacional de Saúde (PNS) 2013. A concordância foi estimada como o percentual de casos em que o risco estimado por um escore estava entre 0,8 e 1,25 vezes o risco estimado pelo outro, e como o coeficiente de Gwet (AC1) para o risco estratificado em baixo, intermediário e alto.
Resultados
O estudo incluiu 4.416 pessoas, entre as 8.952 do componente laboratorial da PNS. O risco mediano (intervalo interquartis) foi de 9,2% (5,1 a 17,8) pelo Framingham, 3,6% (1,7 a 8,2) pelo PCE, e 4,7% (2,8 a 8,1) pelo Globorisk-LAC. O risco estimado pelo Framingham concordou com o Globorisk-LAC em 6,4% dos casos e com o PCE em 1,8%; o PCE concordou com o Globorisk-LAC em 34,7%. Considerando a estratificação do risco, o AC1 foi respectivamente 0,454, 0,489 e 0,874.
Conclusões
Os diferentes escores de risco cardiovascular concordam muito pouco entre si. Os motivos para essa discordância fazem do Globorisk-LAC uma ótima opção para suceder o Framingham nas diretrizes brasileiras de dislipidemia.
Palavras-chave:Fatores de Risco de Doenças CardíacasBrasilMedição de RiscoIntrodução
A estratificação do risco cardiovascular é amplamente recomendada por diretrizes clínicas1-5 como uma forma de dosar a abordagem medicamentosa da dislipidemia e outros fatores de risco. Tipicamente, ela começa alocando pessoas com doença cardiovascular a níveis de risco mais elevados. No caso das pessoas sem doença estabelecida, a estratificação envolve aplicar pontos de corte clinicamente relevantes a uma estimativa de risco cardiovascular (probabilidade de um evento numa escala de tempo de dez anos), que por sua vez é obtida através de algum escore prognóstico.
Assim como os demais modelos prognósticos, os escores de risco cardiovascular devem ter sua calibração e discriminação validadas na população onde serão usados. No estudoProspective Urban Rural Epidemiology (PURE), por exemplo, os eventos cardiovasculares foram mais frequentes nos países de baixa e média renda, mesmo eles tendo escores menores do que os de alta renda.6 Devido a essa situação, os escores são frequentemente recalibrados para as populações onde serão usados.6,7 De fato, a facilidade de recalibra-los para os países-alvo é um motivo importante para a recomendação de determinados escores em algumas diretrizes.1,3
As mais recentes diretrizes brasileiras de dislipidemia4,8 recomendam o escore de risco cardiovascular global de Framingham.9 Já as diretrizes americanas recomendam o Pooled Cohort Equations (PCE),5 que uniu Framingham a outras coortes do mesmo país, recalibrando o resultado final para a população contemporânea dos Estados Unidos.10 Infelizmente, nem o escore de Framingham, nem o PCE foi recalibrado para a população brasileira.11 Os únicos escores de risco cardiovascular calibrados para o Brasil são os da Organização Mundial da Saúde (OMS),12 o Globorisk,13,14 e o Globorisk-LAC.15
Em uma pesquisa brasileira sem representatividade populacional,16 todos os escores examinados superestimaram o risco cardiovascular (inclusive OMS e Globorisk-LAC), apesar de manterem a capacidade de discriminação semelhante àquela encontrada em suas validações iniciais. Dentre os dois escores, o da OMS tem a vantagem de ter superestimado menos o risco cardiovascular naquele estudo. Enquanto isso, o Globorisk-LAC tem o diferencial de ter sido derivado de coortes latino-americanas,17 e sua versão calibrada para os países teve um desempenho semelhante, e possivelmente um pouco melhor do que o escore da OMS quando aplicados às mesmas coortes de base populacional.15
Como nenhum estudo até o momento avaliou a concordância entre o Globorisk-LAC e outros escores de risco cardiovascular, fica a dúvida de quanta diferença faria trocar o Framingham pelo Globorisk-LAC nas diretrizes brasileiras de dislipidemia.4,8 O objetivo deste estudo foi, então, avaliar a concordância entre os escores de Framingham, PCE e Globorisk-LAC na estimação do risco cardiovascular para a população do Brasil.
MétodosDelineamento
Este estudo descritivo de delineamento transversal utilizou dados da Pesquisa Nacional de Saúde (PNS) 2013.18 Conduzida entre agosto e dezembro do referido ano, a pesquisa foi um inquérito domiciliar realizado em três etapas. Os conglomerados (setores censitários) foram selecionados aleatoriamente, com probabilidade proporcional ao número de domicílios particulares permanentes. Dentro destes, os domicílios (dentro dos conglomerados) e um morador com 18 anos de idade ou mais (dentro de cada domicílio) foram selecionados por amostragem aleatória simples.
Ao contrário de sua edição de 2019, a PNS 2013 envolveu a coleta de sangue e urina para exames de laboratório.19 Um quarto dos conglomerados da PNS 2013 foi selecionado para a coleta de exames laboratoriais, com probabilidade inversamente proporcional à dificuldade de coleta. Como a taxa de não resposta excedeu substancialmente os 20% esperados, os dados foram reponderados para serem representativos em nível macrorregional.19
Variáveis
Neste estudo, calculou-se o risco cardiovascular em dez anos conforme o escore de risco cardiovascular global de Framingham,9 PCE (revisão de 2018)10 e Globorisk-LAC15 usando, respectivamente, os pacotes {CVrisk} 1.1.1,20 {PooledCohort} 0.0.221 e {globorisk} 1.0.222 para o ambiente de computação estatística R 4.4.0.23 Sexo, idade, pressão arterial, colesterol total e colesterol HDL (lipoproteína de alta densidade) foram obtidos das respectivas variáveis. Consideram-se negras as pessoas que relataram pele parda ou preta, e fumantes as pessoas que responderam "sim, diariamente" ou "sim, menos que diariamente".
Como a questão sobre o uso de medicação hipertensiva nas últimas duas semanas só foi feita às pessoas que informaram ter hipertensão, imputou-se a resposta "não" àquelas que descreveram não ter o diagnóstico (fora da gravidez) ou nunca terem tido a pressão arterial medida. Da mesma forma, considerou-se como não tendo diabetes mellitus as pessoas que relataram nunca ter tido sua glicemia medida. Depois desse ajuste, imputou-se o diagnóstico de diabetes mellitus às pessoas cuja hemoglobina glicada na PNS 2013 foi igual ou maior a 6,5%, mesmo se essas não tivessem diagnósticos prévios. Por fim, considerou-se como tendo doença cardiovascular as pessoas que relataram já ter tido diagnóstico de qualquer doença cardíaca ou de acidente vascular cerebral.
Amostra
O presente estudo incluiu pessoas com idade entre 40 e 74 anos, sem doença cardiovascular conhecida, e sem valores atípicos nem de pressão arterial sistólica, nem de colesterol total (como especificado no desenvolvimento do Globorisk-LAC).15 Para garantir a validade das comparações, excluímos as pessoas com valores faltantes que impossibilitassem a estimação do risco cardiovascular por qualquer um dos três escores, mesmo quando era possível calcular os demais escores. O peso amostral foi considerado no cálculo de todas as estatísticas, exceto da frequência absoluta.
Análise estatística
Em sua forma contínua, o risco cardiovascular teve sua distribuição descrita por medianas e intervalos interquartis (IIQ). Para descrever a proporção de participantes nos quais um escore concordava com o outro, considerou-se os escores como concordantes quando o risco mais elevado era menos de 25% superior ao risco menos elevado. Essa forma de comparar as estimativas de risco equivale à diferença entre logaritmos proposta por Bland e Altmann,24 só que numa escala mais intuitiva. O ponto de corte de 25% (uma razão de 1,25 entre duas taxas) é uma proposta da metodologia Grading of Recommendations Assessment, Development and Evaluation (GRADE) quando não há um consenso sobre qual é o menor tamanho de efeito clinicamente relevante.25
O risco cardiovascular também foi estratificado em baixo (menos de 10%), intermediário (10% a menos de 20%) e alto (20% ou mais). Como a estratificação do risco cardiovascular varia entre as diretrizes, optou-se pelo ponto de corte de 20% para risco elevado, com base nas diretrizes da OMS1 e da American Heart Association (AHA) / American College of Cardiology (ACC),5 e pelo ponto de corte de 10%, porque corresponde a uma recomendação forte de prescrição de estatina, enquanto a recomendação a partir de 7,5% é condicional.26
A distribuição da população brasileira entre os três estratos foi descrita usando frequência relativa (percentuais). Calculou-se a concordância usando tanto o percentual observado de concordância quanto o coeficiente de concordância de primeira ordem de Gwet (AC1).27 Esse coeficiente tem valores entre −1 e +1, como o coeficiente κ de Cohen, e é mais resistente a paradoxos conhecidos.27,28 Em ambos os casos, desconsiderou-se a natureza ordinal da estratificação do risco cardiovascular, de forma que estratos de risco cardiovascular adjacentes foram considerados discordantes, e não parcialmente concordantes.
Como o pacote estatístico {irrCAC}29 não considera pesos amostrais ao computar o coeficiente, foi desenvolvido neste estudo código de análise para esse fim. A título de análise de sensibilidade, repetiu-se a estimativa do AC1 usando os prontos de corte da AHA/ACC 20185 (5%, 7,5% e 20%), OMS1 (10%, 20% e 30%) e European Society of Cardiology (ESC) / European Atherosclerosis Society (EAS)3 (10%, 15% e 30% para eventos fatais + não-fatais, obtidos multiplicando-se por três os pontos de corte de 3%, 5% e 10% para eventos fatais). Todo o código de análise está disponível abertamente.30
Os gráficos foram feitos com os pacotes {ggplot2} 3.5.131,32 e {ggalluvial} 0.12.5.33,34 Para garantir o entendimento em caso de discromatopsia (ou de impressão em tons de cinza), os gráficos foram feitos usando as escalas de cores de Okabe-Ito,35 disponível no R; "dark2" de Brewer,36 através do pacote {RColorBrewer} 1.1;37 e "cividis",38 através do {viridisLite} 0.4.2.39,40
Considerações éticas
A PNS 2013 foi aprovada pelo Conselho Nacional de Ética em Pesquisa (relatório número 328.159). Os participantes assinaram um termo de consentimento antes de a pesquisa ser realizada, e ao final receberam o resultado dos exames laboratoriais realizados.
Resultados
Dentre os 8.952 participantes do componente laboratorial da PNS 2013, 4.989 (54,0%) tinham entre 40 e 74 anos de idade, 4.781 (51,7%) também tinham dados completos para a estimação do risco cardiovascular e 4.416 (47,0%) não tinham história de doença cardíaca, nem de acidente vascular cerebral. Todos esses 4.416 participantes foram incluídos neste estudo, porque nenhum deles tinha valores atípicos de pressão arterial sistólica ou colesterolemia total.
Os participantes (Tabela 1) eram em sua maioria do sexo feminino (2529; 52,3%), a idade média ± DP foi 53,4 ± 9,4 anos, e praticamente metade (2634; 48,8%) era negra (preta ou parda). Com relação aos fatores clinicamente modificáveis, 796 (17,7%) eram fumantes, 1039 (23,4%) estavam usando anti-hipertensivo e 565 (13,9%) tinham diabetes.
Características demográficas e fatores de risco
Variável
Mediana
Intervalo interquartis
Idade (anos)
52,0
25,0 – 61,0
Colesterol total (mg/dL)
191,0
169,0 – 216,0
Colesterol HDL (mg/dL)
44,0
37,0 –53,0
Pressão arterial sistólica (mmHg)
128,0
116,0 – 140,5
O risco cardiovascular em dez anos teve mediana de 9,2% (IIQ de 5,1 a 17,8) quando estimado pelo escore de Framingham, 3,6% (IIQ de 1,7 a 8,2) pelo PCE, e 4,7% (IIQ de 2,8 a 8,1) pelo Globorisk-LAC (Figura 1). O risco foi considerado baixo em 52,8% dos casos quando estimado pelo escore de Framingham, 79,6% pelo PCE, e 82,2% pelo Globorisk-LAC. Adicionalmente, o risco foi considerado intermediário em 20,8% pelo Framingham, 6,5% pelo PCE, e 3,3% pelo Globorisk-LAC.
Risco cardiovascular estimado para 10 anos, conforme o escore utilizado para a estimativa.
Considerando-se o risco cardiovascular numa escala contínua, o escore de Framingham concordou com o Globorisk-LAC na estimação do risco cardiovascular em apenas 6,4% dos casos, e com o PCE em apenas 1,8% (Figura 2). Enquanto isso, o escore PCE concordou com o Globorisk-LAC em 34,7% dos casos. O risco estimado pelo Framingham foi substancialmente maior do que aquele estimado pelo Globorisk-LAC em 93,5% dos casos, e do que aquele estimado pelo PCE em 98,2% dos casos. Enquanto isso, o risco estimado pelo PCE foi substancialmente menor do que o estimado pelo Globorisk-LAC em 50,8% dos casos.
Razão entre o risco cardiovascular estimado para 10 anos entre diferentes escores.
Estratificando-se o risco cardiovascular em baixo, intermediário e alto (Figura Central), o escore de Framingham concordou com o Globorisk-LAC em 58,8% dos casos (coeficiente de concordância AC1 de 0,454) e com o PCE em 71,6% (AC1 de 0,489). Enquanto isso, o escore PCE concordou com o Globorisk-LAC em 89,5% dos casos (AC1 de 0,874). Observa-se na Figura Central que a maioria dos casos considerados de alto risco pelo Framingham são considerados de risco intermediário pelos outros dois escores. Da mesma forma, a maioria dos casos classificados como de risco intermediário pelo Framingham são considerados de risco baixo pelos outros dois escores.
Concordância entre Framingham, Pooled Cohort Equations e Globorisk-LAC da estimação do risco cardiovascular no Brasil, 2013
Diagrama aluvial da concordância entre diferentes escores na estratificação do risco cardiovascular para 10 anos.
A comparação entre os escores PCE e Globorisk-LAC na estratificação do risco cardiovascular (Figura Central) difere do que se poderia esperar pela comparação entre os valores dos riscos estimados para dez anos (Figura 2). Mais especificamente, pouco mais da metade dos casos classificados como de alto risco pelo primeiro escore foi de risco intermediário pelo segundo, e um terço dos classificados como de risco intermediário pelo primeiro foi de risco baixo pelo segundo (Figura Central).
Enquanto isso, o risco estimado pelo PCE foi substancialmente menor do que aquele estimado pelo Globorisk-LAC em metade dos casos, como demonstrado na Figura 2, sugerindo que, em comparação ao Globorisk-LAC, o PCE esteja subestimando o risco nos casos de menor risco, e superestimando nos casos de maior risco. De fato, observa-se na Figura 1 que os valores de risco estimados pelo PCE têm uma distribuição mais ampla do que aqueles estimados pelo Globorisk-LAC.
Na análise de sensibilidade, os coeficientes de concordância AC1 diminuíram substancialmente com a estratificação do risco da AHA/ACC 2018: 0,191 entre Framingham e Globorisk-LAC, 0,228 entre Framingham e PCE e 0.704 entre PCE e Globorisk-LAC. Por outro lado, os coeficientes praticamente não se alteraram com a estratificação pela OMS (0,476, 0,490 e 0.870, respectivamente) ou pela ESC/EAS 2019 (0,464, 0,486 e 0,848, respectivamente).
Discussão
Os três escores examinados concordam muito pouco entre si na estimação do risco cardiovascular na população brasileira. Os dois escores mais parecidos entre si foram o PCE e o Globorisk-LAC, com um coeficiente de concordância de 0,874. Em comparação ao Globorisk-LAC, o PCE subestima o risco cardiovascular das pessoas com menor risco e superestima o risco daquelas com maior risco, causando neste último caso uma mudança na estratificação do risco cardiovascular. Quando comparado a esses dois escores, o de Framingham superestima substancialmente o risco de mais de 95% da população, resultando num coeficiente de concordância de 0,489 e 0,454, respectivamente.
Os mesmos dados foram estudados por Malta et al.,41 que encontraram uma prevalência de 44% de risco cardiovascular elevado pelo critério da AHA em 2013;42 38% pelo critério da Sociedade Brasileira de Cardiologia (SBC); e 19% pelo critério do Framingham. O percentual observado de concordância entre o critério da SBC e o da AHA foi de 43% para alto risco e 55% para baixo risco, e entre SBC e Framingham foi 51% para alto risco e 100% para baixo risco.
Em comparação, o presente estudo observou um percentual de 59% de concordância entre Framingham e PCE, considerando simultaneamente três estratos de risco cardiovascular definidos pelos mesmos pontos de corte (< 10%, 10–20%, ≥ 20%) aplicados a todos os escores. Além disso, no presente estudo utilizou-se a versão mais recente, de 2018, do PCE.10 Por fim, o foco do presente estudo foi a comparação entre os escores de risco cardiovascular, enquanto Malta et al. focaram na falta de intercambialidade entre os critérios.
Um dos poucos estudos a comparar o Globorisk com outros escores foi o de Osei-Yeboah et al.,43 com uma prevalência de risco cardiovascular elevado de 13% pelo Framingham (ponto de corte: 20% em dez anos), 0% pelo Globorisk (ponto de corte: 30%) e 3% pelo PCE2018 (ponto de corte: 20%). O coeficiente κ de Cohen foi 0,008 para a concordância entre Framingham e Globorisk, 0,432 entre Framingham e PCE, e 0,020 entre PCE e Globorisk. Outro estudo44 comparou o Globorisk (19% de risco cardiovascular alto ou muito alto) com Framingham (37% de risco alto ou muito alto) em ambulatórios de diabetes mellitus em hospitais de Bangladesh, chegando a um κ de 0,34.
No presente estudo, a concordância entre Framingham e PCE (AC1 de 0,489) foi próxima àquela estimada estudo de Osei-Yeboah et al.43 enquanto a concordância do Globorisk-LAC com Framingham (AC1 de 0,454) e PCE (AC1 de 0,874) foi muito superior. A diferença se deve, ao menos em parte, ao estudo de Osei-Yeboah et al.43 ter adotado um ponto de corte mais elevado (30%) para considerar alto o risco cardiovascular estimado pelo Globorisk, resultando em nenhum dos quase 14 mil participantes ter sido considerado de alto risco pelo Globorisk. Já o outro estudo44 usou os mesmos pontos de corte do presente estudo, resultando em uma estimativa mais próxima à nossa para a concordância entre Globorisk e Framingham.
Note-se que ambos os estudos43,44 usaram a versão mundial do Globorisk,13,14 derivada principalmente de coortes conduzidas em países de alta renda. Até onde sabemos, o presente estudo é o primeiro a estimar a concordância de outros escores com o Globorisk-LAC,15,17 que foi derivado de coortes conduzidas na América Latina e Caribe.
Vários estudos compararam o escore de risco de Framingham ao PCE na estratificação do risco cardiovascular. Em estudos realizados em ambulatórios,45-48 a prevalência de risco cardiovascular elevado variou de 2% a 50% pelo escore de Framingham e de 5% a 50% pelo PCE, e o coeficiente κ de Cohen variou de 0,049 a 0,745. Já os estudos de base populacional43,49-51 encontraram prevalências de risco elevado que variavam de 3% a 19% pelo Framingham e de menos de 1% até 10% pelo PCE, e valores de κ entre 0,29 e 0,55.
Tanto nos estudos com pacientes, quanto naqueles com base populacional, o coeficiente κ de Cohen foi menor nas pesquisas com menores prevalências de risco cardiovascular elevado. Isso ilustra o viés do κ, que subestima a concordância quando uma das categorias tem baixa prevalência.27,28 Esse é o motivo pelo qual o presente estudo optou pelo coeficiente AC1 de Gwet27 como estimando para a concordância na estratificação do risco cardiovascular.
A presente pesquisa foi a primeira a comparar os três escores de risco cardiovascular (Framingham, PCE e Globorisk-LAC) usando o método de Bland e Altmann.24 Ao menos um outro estudo já usou esse método no contexto do risco cardiovascular, mas para comparar o escore Globorisk com o da OMS.52 Diferentemente daquele estudo, optou-se por trabalhar com a razão (em vez da diferença) entre os riscos, e compará-la com um intervalo predefinido (0,8 a 1,25), de forma a poder estimar em que percentual dos casos os escores concordam entre si.
Antes de prosseguir com a interpretação dos achados e a discussão de suas implicações, cumpre ressaltar as limitações deste estudo. O banco de dados laboratoriais da PNS 2013 não contém a variável do conglomerado, impossibilitando que o delineamento da amostragem seja considerado em qualquer forma de inferência estatística, tais como valores p ou intervalos de confiança. O presente estudo usou esses dados mesmo assim, porque a PNS 2013 é o mais recente inquérito nacionalmente representativo que coletou os dados necessários para estimar o risco cardiovascular usando a versão laboratorial dos escores.
Ainda que a baixa taxa de resposta do componente laboratorial tenha o potencial de enviesar as estimativas, espera-se que a reponderação tenha minimizado esse potencial viés. Cumpre lembrar ainda que o presente estudo não verificou a ocorrência dos desfechos preditos, de forma que sua contribuição se limita a comparar os escores com relação à concordância mútua, sem a pretensão de informar qual escore tem a melhor calibração e/ou discriminação.
Por fim, no intuito de manter um número razoável de comparações, este estudo incluiu apenas três escores de risco cardiovascular, deixando de fora, em especial, o escore da OMS e o recém-proposto PREVENT. Como o escore da OMS12 faz parte da estratégia HEARTS,53 e teve um desempenho superior no estudo de Camargos et al.,16 justifica-se aprofundar seu estudo na população brasileira. No caso do PREVENT, com suas inovações conceituais54 e desempenho exemplar,55 seria necessário recalibrá-lo para a população brasileira antes de cogitar sua adoção. Pode valer a pena, já que nos Estados Unidos, país para o qual foi calibrado, o novo escore discorda do PCE para vários milhões de pessoas.56
Cabe agora considerar os motivos pelos quais os diferentes escores de risco cardiovascular discordam tanto entre si quando aplicados à população brasileira. Um motivo é mais óbvio: diferentes escores foram derivados de diferentes coortes e calibrados para diferentes populações. Isso é grande parte do motivo de utilizar-se nos Estados Unidos hoje o PCE, e não o Framingham.42 Também é o motivo deste estudo investigar as consequências de uma eventual substituição do Framingham pelo Globorisk-LAC nas diretrizes brasileiras de dislipidemia.4,8
Outro motivo para a discordância entre os escores são as diferenças na composição dos eventos cardiovasculares preditos. Tanto o PCE quanto o Globorisk-LAC focaram nos eventos ateroscleróticos duros: infarto do miocárdio não fatal, morte coronariana e AVE isquêmico ou hemorrágico.10,15,42 Já no escore de Framingham, o desfecho inclui ainda insuficiência coronariana, angina, ataque isquêmico transitório, doença arterial periférica e insuficiência cardíaca.9
Esses motivos têm consequências para uma eventual atualização das diretrizes clínicas,4,8 no que diz respeito à adaptação de recomendações internacionais. Por exemplo, hoje se recomenda nos Estados Unidos a profilaxia medicamentosa da doença aterosclerótica a partir de 7,5%5 ou 10%26 de risco cardiovascular em dez anos, estimado através do PCE. No que diz respeito à interpretação dos escores de risco cardiovascular, os mesmos pontos de corte poderiam ser usados no Brasil se o risco for estimado pelo Globorisk-LAC. Não se pode dizer o mesmo sobre o escore de Framingham, cuja publicação9 não traz uma versão específica para doença aterosclerótica. Ademais, cumpre lembrar que o escore de Framingham nunca foi recalibrado para a população brasileira.11
Conclusão
Na população brasileira, o risco cardiovascular estimado para dez anos difere substancialmente entre os escores de risco de Framingham, do PCE e do Globorisk-LAC. A adequação à realidade brasileira e a similaridade de desfechos com PCE fazem do Globorisk-LAC uma ótima opção para suceder o Framingham na próxima atualização das diretrizes brasileiras de dislipidemia e prevenção de aterosclerose, bem como dos respectivos protocolos clínicos e diretrizes terapêuticas.
Os conteúdos estão disponíveis no DOI https://doi.org/10.5281/zenodo.13948103.
O presente estudo não teve fontes de financiamento externas.
Não há vinculação deste estudo a programas de pós-graduação.
Este estudo foi aprovado pela Comissão Nacional de Ética em Pesquisa sob o número de protocolo 328.159. Todos os procedimentos envolvidos nesse estudo estão de acordo com a Declaração de Helsinki de 1975, atualizada em 2013. O consentimento informado foi obtido de todos os participantes incluídos no estudo.
Os autores não utilizaram ferramentas de inteligência artificial no desenvolvimento deste trabalho.
ReferênciasWorld Health OrganizationPrevention of Cardiovascular Disease: Guidelines for Assessment and Management of Total Cardiovascular Risk [Internet]GenevaWorld Health Organization2007citaed 2025 May 9Available from: https://apps.who.int/iris/handle/10665/436851 World Health Organization. Prevention of Cardiovascular Disease: Guidelines for Assessment and Management of Total Cardiovascular Risk [Internet]. Geneva: World Health Organization; 2007 [cited 2025 May 9]. Available from: https://apps.who.int/iris/handle/10665/43685ManciaGKreutzRBrunströmMBurnierMGrassiGJanuszewiczA2023 ESH Guidelines for the Management of Arterial Hypertension The Task Force for the Management of Arterial Hypertension of the European Society of Hypertension: Endorsed by the International Society of Hypertension (ISH) and the European Renal Association (ERA)J Hypertens202341121874207110.1097/HJH.00000000000034802 Mancia G, Kreutz R, Brunström M, Burnier M, Grassi G, Januszewicz A, et al. 2023 ESH Guidelines for the Management of Arterial Hypertension The Task Force for the Management of Arterial Hypertension of the European Society of Hypertension: Endorsed by the International Society of Hypertension (ISH) and the European Renal Association (ERA). J Hypertens. 2023;41(12):1874-2071. doi: 10.1097/HJH.0000000000003480.MachFBaigentCCatapanoALKoskinasKCCasulaMBadimonL2019 ESC/EAS Guidelines for the Management of Dyslipidaemias: Lipid Modification to Reduce Cardiovascular RiskEur Heart J202041111118810.1093/eurheartj/ehz4553 Mach F, Baigent C, Catapano AL, Koskinas KC, Casula M, Badimon L, et al. 2019 ESC/EAS Guidelines for the Management of Dyslipidaemias: Lipid Modification to Reduce Cardiovascular Risk. Eur Heart J. 2020;41(1):111-88. doi: 10.1093/eurheartj/ehz455.FaludiAAIzarMCOSaraivaJFKChacraAPMBiancoHTAfiuneANetoAtualização da Diretriz Brasileira de Dislipidemias e Prevenção da Aterosclerose – 2017Arq Bras Cardiol20171092 Suppl 117610.5935/abc.201701214 Faludi AA, Izar MCO, Saraiva JFK, Chacra APM, Bianco HT, Afiune A Neto, et al. Atualização da Diretriz Brasileira de Dislipidemias e Prevenção da Aterosclerose – 2017. Arq Bras Cardiol. 2017;109(2 Suppl 1):1-76. doi: 10.5935/abc.20170121.GrundySMStoneNJBaileyALBeamCBirtcherKKBlumenthalRS2018 AHA/ACC/AACVPR/AAPA/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA Guideline on the Management of Blood Cholesterol: Executive Summary: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice GuidelinesCirculation2019139251046108110.1161/CIR.00000000000006245 Grundy SM, Stone NJ, Bailey AL, Beam C, Birtcher KK, Blumenthal RS, et al. 2018 AHA/ACC/AACVPR/AAPA/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA Guideline on the Management of Blood Cholesterol: Executive Summary: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. Circulation. 2019;139(25):1046-81. doi: 10.1161/CIR.0000000000000624.JosephPYusufSLeeSFIbrahimQTeoKRangarajanSPrognostic Validation of a Non-Laboratory and a Laboratory Based Cardiovascular Disease Risk Score in Multiple Regions of the WorldHeart2018104758158710.1136/heartjnl-2017-3116096 Joseph P, Yusuf S, Lee SF, Ibrahim Q, Teo K, Rangarajan S, et al. Prognostic Validation of a Non-Laboratory and a Laboratory Based Cardiovascular Disease Risk Score in Multiple Regions of the World. Heart. 2018;104(7):581-7. doi: 10.1136/heartjnl-2017-311609.GulayinPEDanaeiGGutierrezLPoggioRPonzoJLanasFExternal Validation of Cardiovascular Risk Scores in the Southern Cone of Latin America: Which Predicts Better?Rev Argent Cardiol201886111510.7775/rac.es.v86.i1.129087 Gulayin PE, Danaei G, Gutierrez L, Poggio R, Ponzo J, Lanas F, et al. External Validation of Cardiovascular Risk Scores in the Southern Cone of Latin America: Which Predicts Better? Rev Argent Cardiol. 2018;86(1):1-15. doi: 10.7775/rac.es.v86.i1.12908.Brasil. Ministério da SaúdeProtocolo Clínico e Diretrizes Terapêuticas da Dislipidemia: Prevenção de Eventos Cardiovasculares e PancreatiteBrasíliaMinistério da Saúde20208 Brasil. Ministério da Saúde. Protocolo Clínico e Diretrizes Terapêuticas da Dislipidemia: Prevenção de Eventos Cardiovasculares e Pancreatite. Brasília: Ministério da Saúde; 2020.D'AgostinoRBSrVasanRSPencinaMJWolfPACobainMMassaroJMGeneral Cardiovascular Risk Profile for Use in Primary care: The Framingham Heart StudyCirculation2008117674375310.1161/CIRCULATIONAHA.107.6995799 D'Agostino RB Sr, Vasan RS, Pencina MJ, Wolf PA, Cobain M, Massaro JM, et al. General Cardiovascular Risk Profile for Use in Primary care: The Framingham Heart Study. Circulation. 2008;117(6):743-53. doi: 10.1161/CIRCULATIONAHA.107.699579.YadlowskySHaywardRASussmanJBMcClellandRLMinYIBasuSClinical Implications of Revised Pooled Cohort Equations for Estimating Atherosclerotic Cardiovascular Disease RiskAnn Intern Med20181691202910.7326/M17-301110 Yadlowsky S, Hayward RA, Sussman JB, McClelland RL, Min YI, Basu S. Clinical Implications of Revised Pooled Cohort Equations for Estimating Atherosclerotic Cardiovascular Disease Risk. Ann Intern Med. 2018;169(1):20-9. doi: 10.7326/M17-3011.Carrillo-LarcoRMAltez-FernandezCPacheco-BarriosNBambsCIrazolaVMirandaJJCardiovascular Disease Prognostic Models in Latin America and the Caribbean: A Systematic ReviewGlob Heart2019141819310.1016/j.gheart.2019.03.00111 Carrillo-Larco RM, Altez-Fernandez C, Pacheco-Barrios N, Bambs C, Irazola V, Miranda JJ, et al. Cardiovascular Disease Prognostic Models in Latin America and the Caribbean: A Systematic Review. Glob Heart. 2019;14(1):81-93. doi: 10.1016/j.gheart.2019.03.001.WHO CVD Risk Chart Working GroupWorld Health Organization Cardiovascular Disease Risk Charts: Revised Models to Estimate Risk in 21 Global RegionsLancet Glob Health20197101332134510.1016/S2214-109X(19)30318-312 WHO CVD Risk Chart Working Group. World Health Organization Cardiovascular Disease Risk Charts: Revised Models to Estimate Risk in 21 Global Regions. Lancet Glob Health. 2019;7(10):1332-45. doi: 10.1016/S2214-109X(19)30318-3.HajifathalianKUedaPLuYWoodwardMAhmadvandAAguilar-SalinasCAA Novel Risk Score to Predict Cardiovascular Disease Risk in National Populations (Globorisk): A Pooled Analysis of Prospective Cohorts and Health Examination SurveysLancet Diabetes Endocrinol20153533935510.1016/S2213-8587(15)00081-913 Hajifathalian K, Ueda P, Lu Y, Woodward M, Ahmadvand A, Aguilar-Salinas CA, et al. A Novel Risk Score to Predict Cardiovascular Disease Risk in National Populations (Globorisk): A Pooled Analysis of Prospective Cohorts and Health Examination Surveys. Lancet Diabetes Endocrinol. 2015;3(5):339-55. doi: 10.1016/S2213-8587(15)00081-9.UedaPWoodwardMLuYHajifathalianKAl-WotayanRAguilar-SalinasCALaboratory-Based and Office-Based Risk Scores and Charts to Predict 10-Year Risk of Cardiovascular Disease in 182 Countries: A Pooled Analysis of Prospective Cohorts and Health SurveysLancet Diabetes Endocrinol20175319621310.1016/S2213-8587(17)30015-314 Ueda P, Woodward M, Lu Y, Hajifathalian K, Al-Wotayan R, Aguilar-Salinas CA, et al. Laboratory-Based and Office-Based Risk Scores and Charts to Predict 10-Year Risk of Cardiovascular Disease in 182 Countries: A Pooled Analysis of Prospective Cohorts and Health Surveys. Lancet Diabetes Endocrinol. 2017;5(3):196-213. doi: 10.1016/S2213-8587(17)30015-3.Cohorts Consortium of Latin America and the Caribbean (CC-LAC)Derivation, Internal Validation, and Recalibration of a Cardiovascular Risk Score for Latin America and the Caribbean (Globorisk-LAC): A Pooled Analysis of Cohort StudiesLancet Reg Health Am2022911210.1016/j.lana.2022.10025815 Cohorts Consortium of Latin America and the Caribbean (CC-LAC). Derivation, Internal Validation, and Recalibration of a Cardiovascular Risk Score for Latin America and the Caribbean (Globorisk-LAC): A Pooled Analysis of Cohort Studies. Lancet Reg Health Am. 2022;9:1-12. doi: 10.1016/j.lana.2022.100258.CamargosAPBarretoSBrantLRibeiroALPDhingraLSAminorroayaAPerformance of Contemporary Cardiovascular Risk Stratification Scores in Brazil: An Evaluation in the ELSA-Brasil StudyOpen Heart2024111e00276210.1136/openhrt-2024-00276216 Camargos AP, Barreto S, Brant L, Ribeiro ALP, Dhingra LS, Aminorroaya A, et al. Performance of Contemporary Cardiovascular Risk Stratification Scores in Brazil: An Evaluation in the ELSA-Brasil Study. Open Heart. 2024;11(1):e002762. doi: 10.1136/openhrt-2024-002762.Cohorts Consortium of Latin America and the Caribbean (CC-LAC)Cohort Profile: The Cohorts Consortium of Latin America and the Caribbean (CC-LAC)Int J Epidemiol202049514371437g10.1093/ije/dyaa07317 Cohorts Consortium of Latin America and the Caribbean (CC-LAC). Cohort Profile: The Cohorts Consortium of Latin America and the Caribbean (CC-LAC). Int J Epidemiol. 2020;49(5):1437-1437g. doi: 10.1093/ije/dyaa073.SzwarcwaldCLMaltaDCPereiraCAVieiraMLCondeWLSouzaPRJrNational Health Survey in Brazil: Design and Methodology of ApplicationCien Saude Colet201419233334210.1590/1413-81232014192.1407201218 Szwarcwald CL, Malta DC, Pereira CA, Vieira ML, Conde WL, Souza PR Jr, et al. National Health Survey in Brazil: Design and Methodology of Application. Cien Saude Colet. 2014;19(2):333-42. doi: 10.1590/1413-81232014192.14072012.SzwarcwaldCLMaltaDCSouzaPRBJrAlmeidaWDSDamacenaGNPereiraCALaboratory Exams of the National Health Survey: Methodology of Sampling, Data Collection and AnalysisRev Bras Epidemiol201922Suppl 2E190004.SUPL.210.1590/1980-549720190004.supl.219 Szwarcwald CL, Malta DC, Souza PRB Jr, Almeida WDS, Damacena GN, Pereira CA, et al. Laboratory Exams of the National Health Survey: Methodology of Sampling, Data Collection and Analysis. Rev Bras Epidemiol. 2019;22(Suppl 2):E190004.SUPL.2. doi: 10.1590/1980-549720190004.supl.2.CastroVCVrisk: Compute Risk Scores for Cardiovascular Diseases [Internet]ViennaViennaR Foundation for Statistical Computing2020citaed 2025 May 9Available from: https://CRAN.R-project.org/package=CVrisk20 Castro V. CVrisk: Compute Risk Scores for Cardiovascular Diseases [Internet]. Vienna: Vienna: R Foundation for Statistical Computing; 2020 [cited 2025 May 9]. Available from: https://CRAN.R-project.org/package=CVriskJaegerBPooledCohort: Predicted Risk for CVD Using Pooled Cohort Equations, PREVENT Equations, and Other Contemporary CVD Risk Calculators. [Internet]ViennaR Foundation for Statistical Computing2021citaed 2025 May 9Available from: https://CRAN.R-project.org/package=PooledCohort21 Jaeger B. PooledCohort: Predicted Risk for CVD Using Pooled Cohort Equations, PREVENT Equations, and Other Contemporary CVD Risk Calculators. [Internet]. Vienna: R Foundation for Statistical Computing; 2021 [cited 2025 May 9]. Available from: https://CRAN.R-project.org/package=PooledCohortBoyerCDanaiGHajifathalianKUedaPLarcoRMCGloborisk: Global CVD Risk Calculator [Internet]2022citaed 2025 May 9Available from: https://github.com/boyercb/globorisk22 Boyer C, Danai G, Hajifathalian K, Ueda P, Larco RMC. Globorisk: Global CVD Risk Calculator [Internet]. 2022 [cited 2025 May 9]. Available from: https://github.com/boyercb/globoriskR Core TeamR: A Language and Environment for Statistical Computing [Internet]ViennaR Foundation for Statistical Computing2024citaed 2025 May 9Available from: https://www.R-project.org/23 R Core Team. R: A Language and Environment for Statistical Computing [Internet]. Vienna: R Foundation for Statistical Computing; 2024 [cited 2025 May 9]. Available from: https://www.R-project.org/BlandJMAltmanDGMeasuring Agreement in Method Comparison StudiesStat Methods Med Res19998213516010.1177/09622802990080020424 Bland JM, Altman DG. Measuring Agreement in Method Comparison Studies. Stat Methods Med Res. 1999;8(2):135-60. doi: 10.1177/096228029900800204.GuyattGHOxmanADKunzRBrozekJAlonso-CoelloPRindDGRADE Guidelines 6Rating the Quality of Evidence--Imprecision. J Clin Epidemiol201164121283129310.1016/j.jclinepi.2011.01.01225 Guyatt GH, Oxman AD, Kunz R, Brozek J, Alonso-Coello P, Rind D, et al. GRADE Guidelines 6. Rating the Quality of Evidence--Imprecision. J Clin Epidemiol. 2011;64(12):1283-93. doi: 10.1016/j.jclinepi.2011.01.012.MangioneCMBarryMJNicholsonWKCabanaMChelmowDCokerTRStatin Use for the Primary Prevention of Cardiovascular Disease in Adults: US Preventive Services Task Force Recommendation StatementJAMA2022328874675310.1001/jama.2022.1304426 Mangione CM, Barry MJ, Nicholson WK, Cabana M, Chelmow D, Coker TR, et al. Statin Use for the Primary Prevention of Cardiovascular Disease in Adults: US Preventive Services Task Force Recommendation Statement. JAMA. 2022;328(8):746-53. doi: 10.1001/jama.2022.13044.GwetKLComputing Inter-Rater Reliability and its Variance in the Presence of High AgreementBr J Math Stat Psychol200861Pt 1294810.1348/000711006X12660027 Gwet KL. Computing Inter-Rater Reliability and its Variance in the Presence of High Agreement. Br J Math Stat Psychol. 2008;61(Pt 1):29-48. doi: 10.1348/000711006X126600.FeinsteinARCicchettiDVHigh Agreement but Low Kappa: IThe Problems of Two Paradoxes. J Clin Epidemiol199043654354910.1016/0895-4356(90)90158-l28 Feinstein AR, Cicchetti DV. High Agreement but Low Kappa: I. The Problems of Two Paradoxes. J Clin Epidemiol. 1990;43(6):543-9. doi: 10.1016/0895-4356(90)90158-l.KilemL. GwetPh.DirrCAC: Computing Chance-Corrected Agreement Coefficients (CAC) [Internet]ViennaR Foundation for Statistical Computing2019citaed 2025 May 9Available from: https://CRAN.R-project.org/package=irrCAC29 Kilem L. Gwet, Ph.D. irrCAC: Computing Chance-Corrected Agreement Coefficients (CAC) [Internet]. Vienna: R Foundation for Statistical Computing; 2019 [cited 2025 May 9]. Available from: https://CRAN.R-project.org/package=irrCACFontenelleLFleofontenelle/rcv_2013: Public release [Internet]GeneveZenodo2024citaed 2025 May 9Available from: https://zenodo.org/doi/10.5281/zenodo.1394810330 Fontenelle LF. leofontenelle/rcv_2013: Public release [Internet]. Geneve: Zenodo; 2024 [cited 2025 May 9]. Available from: https://zenodo.org/doi/10.5281/zenodo.13948103WickhamHGgplot2: Elegant Graphics for Data Analysis [Internet]ChamSpringer International Publishing201631 Wickham H. Ggplot2: Elegant Graphics for Data Analysis [Internet]. Cham: Springer International Publishing; 2016.WickhamHChangWHenryLPedersenTLTakahashiKWilkeCGgplot2: Create Elegant Data Visualisations Using the Grammar of Graphics [Internet]ViennaR Foundation for Statistical Computing2024citaed 2025 May 9Available from: https://CRAN.R-project.org/package=ggplot232 Wickham H, Chang W, Henry L, Pedersen TL, Takahashi K, Wilke C, et al. Ggplot2: Create Elegant Data Visualisations Using the Grammar of Graphics [Internet]. Vienna: R Foundation for Statistical Computing; 2024 [cited 2025 May 9]. Available from: https://CRAN.R-project.org/package=ggplot2BrunsonJCGgalluvial: Layered Grammar for Alluvial PlotsJ Open Source Softw20205492017201710.21105/joss.0201733 Brunson JC. Ggalluvial: Layered Grammar for Alluvial Plots. J Open Source Softw. 2020;5(49):2017. doi: 10.21105/joss.02017.BrunsonJCReadQDGgalluvial: Alluvial Plots in "ggplot2" [Internet]ViennaR Foundation for Statistical Computing2023citaed 2025 May 9Available from: https://CRAN.R-project.org/package=ggalluvial34 Brunson JC, Read QD. Ggalluvial: Alluvial Plots in "ggplot2" [Internet]. Vienna: R Foundation for Statistical Computing; 2023 [cited 2025 May 9]. Available from: https://CRAN.R-project.org/package=ggalluvialOkabeMItoKHow to Make Figures and Presentations that are Friendly to Color Blind People. [Internet]TokyoUniversity of Tokyo2002citaed 2025 May 9Available from: https://jfly.uni-koeln.de/html/color_blind/35 Okabe M, Ito K. How to Make Figures and Presentations that are Friendly to Color Blind People. [Internet]. Tokyo: University of Tokyo; 2002 [cited 2025 May 9]. Available from: https://jfly.uni-koeln.de/html/color_blind/BrewerCAHatchardGWHarrowerMAColorBrewer in Print: A Catalog of Color Schemes for MapsCartogr Geogr Inf Sci200330153210.1559/15230400310001092936 Brewer CA, Hatchard GW, Harrower MA. ColorBrewer in Print: A Catalog of Color Schemes for Maps. Cartogr Geogr Inf Sci. 2003;30(1):5-32. doi: 10.1559/152304003100010929.NeuwirthERColorBrewer: ColorBrewer Palettes [Internet]ViennaR Foundation for Statistical Computing2002citaed 2025 May 9Available from: https://CRAN.R-project.org/package=RColorBrewer37 Neuwirth E. RColorBrewer: ColorBrewer Palettes [Internet]. Vienna: R Foundation for Statistical Computing; 2002 [cited 2025 May 9]. Available from: https://CRAN.R-project.org/package=RColorBrewerNuñezJRAndertonCRRenslowRSOptimizing Colormaps with Consideration for Color Vision Deficiency to Enable Accurate Interpretation of Scientific DataPLoS One2018137e019923910.1371/journal.pone.019923938 Nuñez JR, Anderton CR, Renslow RS. Optimizing Colormaps with Consideration for Color Vision Deficiency to Enable Accurate Interpretation of Scientific Data. PLoS One. 2018;13(7):e0199239. doi: 10.1371/journal.pone.0199239.Sjmgarnier/viridisLite: CRAN release v0.4.2 [Internet]DeneveZenodo2023citaed 2025 May 9Available from: https://zenodo.org/records/789087539 Sjmgarnier/viridisLite: CRAN release v0.4.2 [Internet]. Deneve: Zenodo; 2023 [cited 2025 May 9]. Available from: https://zenodo.org/records/7890875GarnierSviridislite: Colorblind-Friendly Color Maps (Lite Version) [Internet]ViennaR Foundation for Statistical Computing2023citaed 2025 May 9Available from: https://CRAN.R-project.org/package=viridisLite40 Garnier S. viridislite: Colorblind-Friendly Color Maps (Lite Version) [Internet]. Vienna: R Foundation for Statistical Computing; 2023 [cited 2025 May 9]. Available from: https://CRAN.R-project.org/package=viridisLiteMaltaDCPinheiroPCAzeredoRTSantosFMRibeiroALPBrantLCCPrevalence of High Risk for Cardiovascular Disease Among the Brazilian Adult Population, According to Different Risk Calculators: A Comparative StudyCien Saude Colet20212641221123110.1590/1413-81232021264.0159202141 Malta DC, Pinheiro PC, Azeredo RT, Santos FM, Ribeiro ALP, Brant LCC. Prevalence of High Risk for Cardiovascular Disease Among the Brazilian Adult Population, According to Different Risk Calculators: A Comparative Study. Cien Saude Colet. 2021;26(4):1221-31. doi: 10.1590/1413-81232021264.01592021.GoffDCJrLloyd-JonesDMBennettGCoadySD'AgostinoRBGibbonsR2013 ACC/AHA Guideline on the Assessment of Cardiovascular Risk: A Report of the American College of Cardiology/American Heart Association Task Force on Practice GuidelinesCirculation201412925 Suppl 2S49S7310.1161/01.cir.0000437741.48606.9842 Goff DC Jr, Lloyd-Jones DM, Bennett G, Coady S, D'Agostino RB, Gibbons R, et al. 2013 ACC/AHA Guideline on the Assessment of Cardiovascular Risk: A Report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. Circulation. 2014;129(25 Suppl 2):S49-73. doi: 10.1161/01.cir.0000437741.48606.98.Osei-YeboahJvan CharanteEPMKengneAPOwusu-DaboEvan den BornBHGalenkamp-van der PloegHCardiovascular Risk Estimation Based on Country-of-Birth- and Country-of-Residence-Specific Scores Among Migrants in the Netherlands: The HELIUS StudyInt J Environ Res Public Health20232065148514810.3390/ijerph2006514843 Osei-Yeboah J, van Charante EPM, Kengne AP, Owusu-Dabo E, van den Born BH, Galenkamp-van der Ploeg H, et al. Cardiovascular Risk Estimation Based on Country-of-Birth- and Country-of-Residence-Specific Scores Among Migrants in the Netherlands: The HELIUS Study. Int J Environ Res Public Health. 2023;20(6):5148. doi: 10.3390/ijerph20065148.MondalRRituRBBanikPCCardiovascular Risk Assessment Among Type-2 Diabetic Subjects in Selected Areas of Bangladesh: Concordance Among without Cholesterol-Based WHO/ISH, Globorisk, and Framingham Risk Prediction ToolsHeliyon202178e0772810.1016/j.heliyon.2021.e0772844 Mondal R, Ritu RB, Banik PC. Cardiovascular Risk Assessment Among Type-2 Diabetic Subjects in Selected Areas of Bangladesh: Concordance Among without Cholesterol-Based WHO/ISH, Globorisk, and Framingham Risk Prediction Tools. Heliyon. 2021;7(8):e07728. doi: 10.1016/j.heliyon.2021.e07728.PintoNetoLFSDiasFRBressanFFSantosCROComparison of the ACC/AHA and Framingham Algorithms to Assess Cardiovascular Risk in HIV-Infected PatientsBraz J Infect Dis201721657758010.1016/j.bjid.2017.06.00745 Pinto LFS Neto, Dias FR, Bressan FF, Santos CRO. Comparison of the ACC/AHA and Framingham Algorithms to Assess Cardiovascular Risk in HIV-Infected Patients. Braz J Infect Dis. 2017;21(6):577-80. doi: 10.1016/j.bjid.2017.06.007.Mancera-RincónPGiral-GiraldoHERizo-TelloVZBarrera GaravitoECConcordance between the Framingham ATP III SCORE and ACC/AHA 2013 Scales. En a Cohort of Patients in a Fourth Level Hospital in 2015Acta Med Colomb201843419219946 Mancera-Rincón P, Giral-Giraldo HE, Rizo-Tello VZ, Barrera Garavito EC. Concordance between the Framingham ATP III SCORE and ACC/AHA 2013 Scales. En a Cohort of Patients in a Fourth Level Hospital in 2015. Acta Med Colomb. 2018;43(4):192-9.NeemaSSandhuSVasudevanBVendhanDSSinhaATripathyDMThe Use of Screening Tools for Cardiovascular Risk Assessment in Psoriasis - A Case- Control StudyIndian Dermatol Online J2022131465110.4103/idoj.idoj_421_2147 Neema S, Sandhu S, Vasudevan B, Vendhan DS, Sinha A, Tripathy DM, et al. The Use of Screening Tools for Cardiovascular Risk Assessment in Psoriasis - A Case- Control Study. Indian Dermatol Online J. 2022;13(1):46-51. doi: 10.4103/idoj.idoj_421_21.CameloRMCaram-DeelderCDuarteBPMouraMCBCostaNCMCostaIMCardiovascular Risk Scores among Asymptomatic Adults with HaemophiliaArq Bras Cardiol20231209e2023000410.36660/abc.2023000448 Camelo RM, Caram-Deelder C, Duarte BP, Moura MCB, Costa NCM, Costa IM, et al. Cardiovascular Risk Scores among Asymptomatic Adults with Haemophilia. Arq Bras Cardiol. 2023;120(9):e20230004. doi: 10.36660/abc.20230004.BoatengDAgyemangCBeuneEMeeksKSmeethLSchulzeMBCardiovascular Disease Risk Prediction in Sub-Saharan African Populations - Comparative Analysis of Risk Algorithms in the RODAM StudyInt J Cardiol201825431031510.1016/j.ijcard.2017.11.08249 Boateng D, Agyemang C, Beune E, Meeks K, Smeeth L, Schulze MB, et al. Cardiovascular Disease Risk Prediction in Sub-Saharan African Populations - Comparative Analysis of Risk Algorithms in the RODAM Study. Int J Cardiol. 2018;254:310-5. doi: 10.1016/j.ijcard.2017.11.082.WekesahFMMutuaMKBoatengDGrobbeeDEAsikiGKyobutungiCKComparative Performance of Pooled Cohort Equations and Framingham Risk Scores in Cardiovascular Disease Risk Classification in a Slum Setting in Nairobi KenyaInt J Cardiol Heart Vasc20202810052110052110.1016/j.ijcha.2020.10052150 Wekesah FM, Mutua MK, Boateng D, Grobbee DE, Asiki G, Kyobutungi CK, et al. Comparative Performance of Pooled Cohort Equations and Framingham Risk Scores in Cardiovascular Disease Risk Classification in a Slum Setting in Nairobi Kenya. Int J Cardiol Heart Vasc. 2020;28:100521. doi: 10.1016/j.ijcha.2020.100521.WagnerRGCrowtherNJMicklesfieldLKBouaPRNonterahEAMashinyaFEstimating the Burden of Cardiovascular Risk in Community Dwellers Over 40 Years Old in South Africa, Kenya, Burkina Faso and GhanaBMJ Glob Health202161e00349910.1136/bmjgh-2020-00349951 Wagner RG, Crowther NJ, Micklesfield LK, Boua PR, Nonterah EA, Mashinya F, et al. Estimating the Burden of Cardiovascular Risk in Community Dwellers Over 40 Years Old in South Africa, Kenya, Burkina Faso and Ghana. BMJ Glob Health. 2021;6(1):e003499. doi: 10.1136/bmjgh-2020-003499.DehghanARezaeiFAuneDA Comparative Assessment between Globorisk and WHO Cardiovascular Disease Risk Scores: A Population-Based StudySci Rep2023131142291422910.1038/s41598-023-40820-352 Dehghan A, Rezaei F, Aune D. A Comparative Assessment between Globorisk and WHO Cardiovascular Disease Risk Scores: A Population-Based Study. Sci Rep. 2023;13(1):14229. doi: 10.1038/s41598-023-40820-3.OrdunezPTajerCGazianoTRodriguezYARosendeAJaffeMGThe HEARTS App: A Clinical Tool for Cardiovascular Risk and Hypertension Management in Primary Health CareRev Panam Salud Publica202246e1210.26633/RPSP.2022.1253 Ordunez P, Tajer C, Gaziano T, Rodriguez YA, Rosende A, Jaffe MG. The HEARTS App: A Clinical Tool for Cardiovascular Risk and Hypertension Management in Primary Health Care. Rev Panam Salud Publica. 2022;46:e12. doi: 10.26633/RPSP.2022.12.KhanSSCoreshJPencinaMJNdumeleCERangaswamiJChowSLNovel Prediction Equations for Absolute Risk Assessment of Total Cardiovascular Disease Incorporating Cardiovascular-Kidney-Metabolic Health: A Scientific Statement from the American Heart AssociationCirculation2023148241982200410.1161/CIR.000000000000119154 Khan SS, Coresh J, Pencina MJ, Ndumele CE, Rangaswami J, Chow SL, et al. Novel Prediction Equations for Absolute Risk Assessment of Total Cardiovascular Disease Incorporating Cardiovascular-Kidney-Metabolic Health: A Scientific Statement from the American Heart Association. Circulation. 2023;148(24):1982-2004. doi: 10.1161/CIR.0000000000001191.KhanSSMatsushitaKSangYBallewSHGramsMESurapaneniADevelopment and Validation of the American Heart Association's PREVENT EquationsCirculation2024149643044910.1161/CIRCULATIONAHA.123.06762655 Khan SS, Matsushita K, Sang Y, Ballew SH, Grams ME, Surapaneni A, et al. Development and Validation of the American Heart Association's PREVENT Equations. Circulation. 2024;149(6):430-49. doi: 10.1161/CIRCULATIONAHA.123.067626.DiaoJAShiIMurthyVLBuckleyTAPatelCJPiersonEProjected Changes in Statin and Antihypertensive Therapy Eligibility with the AHA PREVENT Cardiovascular Risk EquationsJAMA202433212989100010.1001/jama.2024.1253756 Diao JA, Shi I, Murthy VL, Buckley TA, Patel CJ, Pierson E, et al. Projected Changes in Statin and Antihypertensive Therapy Eligibility with the AHA PREVENT Cardiovascular Risk Equations. JAMA. 2024;332(12):989-1000. doi: 10.1001/jama.2024.12537.10.36660/abc.20240405iOriginal ArticleAgreement between Framingham, Pooled Cohort Equations, and Globorisk-LAC in the Estimation of Cardiovascular Risk in Brazil, 20130000-0003-4064-433XFontenelleLeonardo FerreiraConception and design of the research and Analysis and interpretation of the dataStatistical analysis and Writing of the manuscript1SartiThiago DiasConception and design of the research and Analysis and interpretation of the dataCritical revision of the manuscript for content1QuinteGabriela CalloConception and design of the research and Analysis and interpretation of the dataCritical revision of the manuscript for content2AlmeidaAna Paula Santana CoelhoConception and design of the research and Analysis and interpretation of the dataCritical revision of the manuscript for content10000-0002-0987-368XMillJosé GeraldoConception and design of the research and Analysis and interpretation of the dataCritical revision of the manuscript for content3VitoriaESBrazilUniversidade Federal do Espírito Santo (UFES) Departamento de Medicina Social, Vitoria, ES – BrazilVitoriaESBrazilUniversidade Federal do Espírito Santo (UFES) Programa de Pós-Graduação em Saúde Coletiva, Vitoria, ES – BrazilVitoriaESBrazilUniversidade Federal do Espírito Santo (UFES) Programa de Pós-Graduação em Ciências Fisiológicas, Vitoria, ES – BrazilLeonardo Ferreira Fontenelle • Universidade Federal do Espírito Santo (UFES) – Departamento de Saúde Coletiva – Av. Marechal Campos, 1468. Postal Code 29047-105, Bonfim, Vitoria, ES – Brazil E-mail: leonardof@leonardof.med.br
No potential conflict of interest relevant to this article was reported.
Marcio Bittencourt
AbstractBackground
The Framingham risk score and Pooled Cohort Equations (PCE) have never been recalibrated for the Brazilian population. In contrast, the Globorisk-LAC score was recently derived using a methodology analogous to the PCE and has been recalibrated for Latin American countries.
Objectives
To describe the agreement between the Framingham, PCE, and Globorisk-LAC scores in estimating the 10-year cardiovascular risk in the Brazilian population.
Methods
This cross-sectional study used the three scores to estimate cardiovascular risk in participants aged 40 to 74 years without a history of cardiovascular disease based on data from the 2013 National Health Survey (PNS). The agreement was estimated as (i) the percentage of participants in which the risk estimated by one score was between 0.80 and 1.25 times the risk estimated by another score and (ii) based on the Gwet's agreement coefficient (AC1) according to risk categories (low, intermediate, and high).
Results
A total of 4,416 participants were included from 8,952 participants from the PNS with a laboratory component. The median (interquartile range) of the estimated 10-year cardiovascular risk was 9.2% (5.1 to 17.8) according to the Framingham, 3.6% (1.7 to 8.2) according to the PCE, and 4.7% (2.8 to 8.1) according to the Globorisk-LAC. The risk estimated using the Framingham agreed with the Globorisk-LAC and PCE in 6.4% and 1.8% of the cases, respectively, whereas the PCE and Globorisk-LAC agreed in 34.7% of the cases. When considering the risk stratification, the respective AC1 values were 0.454, 0.489, and 0.874.
Conclusions
The three cardiovascular risk scores showed low levels of agreement with each other. The reasons for this disagreement suggests that Globorisk-LAC is a strong candidate to replace the Framingham in the Brazilian guidelines for dyslipidemia.
Cardiovascular guidelines1-5 widely recommend cardiovascular risk stratification as a tool to guide the pharmacological management of dyslipidemia and other risk factors. Typically, this stratification initiates by assigning participants with established cardiovascular disease to higher-risk categories. For those without established disease, the stratification involves applying clinically relevant cutoff points to the estimated 10-year risk of cardiovascular event obtained from a prognostic risk score.
As with other prognostic models, the calibration and discrimination of cardiovascular risk scores should be validated in the population to which they will be applied. For example, the Prospective Urban Rural Epidemiology study found low-and middle-income countries to have a higher frequency of cardiovascular events, despite having lower risk scores than in high-income countries.6 Because of this, scores are often recalibrated for the populations to which they will be applied.6,7 The ease of recalibrating a given score is important for its recommendation by clinical guidelines.1,3
The recent Brazilian guidelines for dyslipidemia4,8 recommend the Framingham global cardiovascular risk score.9 In contrast, American guidelines recommend the Pooled Cohort Equations (PCE),5 which combine the Framingham with other cohorts from the United States, recalibrating the model to reflect the contemporary population.10 However, the Framingham and PCE have not been recalibrated for the Brazilian population.11 The only cardiovascular risk scores calibrated for the Brazilian population are the World Health Organization (WHO),12 Globorisk,13,14 and Globorisk-LAC.15 In a Brazilian study without representativity of the general population,16 all scores evaluated (including the WHO and Globorisk-LAC) overestimated the cardiovascular risk despite having a discrimination ability similar to that observed in the original validations; among them, the WHO score had a smaller degree of overestimation. In contrast, the Globorisk-LAC has the advantage of being derived from Latin American cohorts,17 and its calibrated version performed similarly or slightly better than the WHO model when applied to the same population-based cohorts.15
As literature lacks studies examining the agreement between the Globorisk-LAC and other cardiovascular risk scores, it remains unclear the impacts of using the Globorisk-LAC instead of the Framingham in the Brazilian guidelines for dyslipidemia.4,8 This study aimed to evaluate the agreement between the Framingham, PCE, and Globorisk-LAC in estimating the 10-year cardiovascular risk in the Brazilian population.
MethodsStudy design
This descriptive cross-sectional study used data from the 2013 National Health Survey (PNS),18 which was a three-stage household survey conducted between August and December of 2013. Census tracts (clusters) were randomly selected using a probability proportional to the number of private permanent households. Households were selected within census tracts, and one resident aged ≥ 18 years within each household, via simple random sampling.
Unlike the 2019 edition, the 2013 PNS included blood and urine for laboratory testing.19 One-quarter of the clusters were chosen for laboratory testing, with a probability inversely proportional to the difficulty of collecting the specimens. As the non-response rate was substantially higher than the expected 20%, data were reweighted to ensure macro-regional representativeness.19
Variables
The 10-year cardiovascular risk was estimated according to the the Framingham global cardiovascular risk score,9 the 2018 PCE revised,10 and the Globorisk-LAC,15 using the {CVrisk} 1.1.1,20 {PooledCohort} 0.0.2,21 and {globorisk} 1.0.222 packages for the R statistical computing environment, version 4.4.0.23 Gender, age, blood pressure, total cholesterol, and high-density lipoprotein cholesterol were derived from the respective survey variables. Participants who self-identified as brown or black were categorized as black. Smokers included participants who answered "yes, daily" or "yes, less than daily". Because the question regarding recent antihypertensive medication use was only asked to the participants who reported a diagnosis of hypertension, a "no" response was imputed for those denying a diagnosis (excluding during pregnancy) or reporting never having their blood pressure measured. Similarly, participants who had never undergone blood glucose testing were classified as not having diabetes mellitus. Subsequently, a diagnosis of diabetes mellitus was imputed for participants with glycated hemoglobin levels ≥ 6.5% in the 2013 PNS, regardless of self-reported diagnosis. Participants who reported a prior diagnosis of any heart disease or stroke were considered to have cardiovascular disease.
Sample
The study included participants aged 40 to 74 without known cardiovascular disease and with systolic blood pressure and total cholesterol values within the range specified by the Globorisk-LAC model.15 To ensure the validity of comparisons, participants with missing values that precluded the risk calculation for any of the three scores were excluded. Sample weights were applied to all statistical analyses, except for absolute frequency.
Statistical analysis
Cardiovascular risk was summarized using medians and interquartile ranges (IQR). The agreement between scores was evaluated by calculating the proportion of participants whose higher estimated risk was less than 25% greater than the lower. This approach is analogous to comparing logarithmic differences, as proposed by Bland & Altmann,24 but presented on a more intuitive scale. The 25% cutoff point (i.e., ratio of 1.25 between two rates) is proposed by the GRADE methodology when there is no consensus regarding the lower clinically relevant effect size.25
The cardiovascular risk was stratified into low (< 10%), intermediate (10% to < 20%), and high (≥ 20%). As the cardiovascular risk stratification varies between guidelines, the 20% threshold for high-risk followed the WHO1 and American Heart Association (AHA)/American College of Cardiology (ACC) guidelines,5 while the 10% threshold followed a strong recommendation for statin prescription, since the 7,5% threshold is a conditional recommendation.26 The distribution of the Brazilian population among the three strata was described using relative frequency (percentage). The agreement was calculated using the observed percentage agreement and Gwet's first-order agreement coefficient (AC1);27 the latter ranges from −1 to +1 (similar to Cohen's κ coefficient) and is more robust to known paradoxes.27,28 In both cases, the ordinal nature of cardiovascular risk stratification was not considered, and adjacent categories were treated as discordant and not partially concordant. As the {irrCAC} package29 does not support sampling weights, custom R code to compute AC1 using sample weights was developed. For the sensitivity analysis, the estimated AC1 was recalculated using cutoff points from the 2018 AHA/ACC5 (5%, 7.5%, and 20%), WHO1 (10%, 20%, and 30%), and European Society of Cardiology (ESC)/European Atherosclerosis Society (EAS)3 (10%, 15%, and 30% for fatal + nonfatal events obtained by multiplying by three the cutoff points of 3%, 5%, and 10% for fatal events). The code used for analysis is openly available.30
Figures were generated using the {ggplot2} 3.5.131,32 and {ggalluvial} 0.12.533,34 packages and the following color palettes designed for accessibility in cases of dyschromatopsia or grayscale printing: Okabe-Ito35 (available from R), "dark2" from Brewer36 ({RColorBrewer} 1.1 package),37 and "cividis"38 ({viridisLite} 0.4.2).39,40
Ethical considerations
The 2013 PNS was approved by the National Research Ethics Committee (number 328,159). The participants signed an informed consent form before participating and received their laboratory test results after the study.
Results
Among the 8,952 participants in the laboratory component of the 2013 PNS, 4,989 (54.0%) were aged between 40 and 74 years, 4,781 (51.7%) had complete data for cardiovascular risk estimation, and 4,416 (47.0%) had no history of heart disease or stroke. All 4,416 participants were included in this study, as none presented atypical values for systolic blood pressure or total cholesterol.
The participants (Table 1) were predominantly female (n = 2,529; 52.3%), with a mean ± SD age of 53.4 ± 9.4 years. Nearly half (n = 2,634; 48.8%) self-identified as black or brown. Regarding modifiable clinical risk factors, 796 (17.7%) were smokers, 1,039 (23.4%) were using antihypertensive medications, and 565 (13.9%) had diabetes.
Demographic characteristics and cardiovascular risk factors of the study population
Variable
Median
Interquartile range
Age (years)
52.0
25.0 to 61.0
Total cholesterol (mg/dL)
191.0
169.0 to 216.0
HDL cholesterol (mg/dL)
44.0
37.0 to 53.0
Systolic blood pressure (mmHg)
128.0
116.0 to 140.5
The median 10-year cardiovascular risk was 9.2% (IQR 5.1% to 17.8%) when estimated using the Framingham, 3.6% (IQR 1.7% to 8.2%) using the PCE, and 4.7% (IQR 2.8% to 8.1%) using the Globorisk-LAC (Figure 1). The risk was considered low in 52.8% of participants on the Framingham, 79.6% on the PCE, and 82.2% on the Globorisk-LAC. An intermediate risk was found in 20.8% of the individuals using the Framingham, 6.5% using the PCE, and 3.3% using Globorisk-LAC.
Estimated 10-year cardiovascular risk across different risk scores.
Considering cardiovascular risk as a continuous variable, the Framingham score agreed with the Globorisk-LAC and PCE in 6.4% and 1.8% of participants, respectively (Figure 2). In contrast, the PCE score agreed with the Globorisk-LAC in 34.7% of the cases. The Framingham produced substantially higher risk estimates than the Globorisk-LAC with each other in 93.5% of participants, and higher than the PCE in 98.2%. Meanwhile, the PCE produced substantially lower risk estimates than the Globorisk-LAC in 50.8% of the participants.
Ratio of estimated 10-year cardiovascular risk between different scores.
When stratifying cardiovascular risk into low, intermediate, and high (Central Illustration), the Framingham agreed with the Globorisk-LAC in 58.8% of the participants (AC1 = 0.454) and with the PCE in 71.6% (AC1 = 0.489). The highest agreement was observed between PCE and Globorisk-LAC (89.5%; AC1 = 0.874). As shown in the Central Illustration, most participants classified as high risk by the Framingham were classified as intermediate risk by the other two scores. Similarly, most participants classified as intermediate risk by the Framingham were categorized as low risk by the other two scores.
Agreement between Framingham, Pooled Cohort Equations, and Globorisk-LAC in the Estimation of Cardiovascular Risk in Brazil, 2013
Alluvial diagram illustrating the agreement between different scores in the 10-year cardiovascular risk stratification.
The comparison between PCE and Globorisk-LAC risk stratification (Central Illustration) deviated from what could be expected by comparing the values of 10-year risk estimates (Figure 2). Specifically, just over half of the participants classified as high risk by the PCE were classified as intermediate risk by Globorisk-LAC, and one-third in the PCE intermediate risk were classified as low risk by the Globorisk-LAC (Central Illustration). However, the PCE produced lower risk estimates than Globorisk-LAC in half of the cases (Figure 2), suggesting that PCE may underestimate the risk in lower-risk participants while overestimating it in those who are higher-risk. Indeed, Figure 1 shows a wider distribution of risk estimates in the PCE compared with the Globorisk-LAC.
In the sensitivity analysis, the AC1 decreased when using the 2018 AHA/ACC risk stratification thresholds: 0.191 between the Framingham and Globorisk-LAC, 0.228 between the Framingham and PCE, and 0.704 between PCE and Globorisk-LAC. In contrast, AC1 values remained relatively stable when using the WHO (0.476, 0.490, and 0.870, respectively) or 2019 ESC/EAS risk stratification (0.464, 0.486, and 0.848, respectively).
Discussion
The three cardiovascular risk scores analyzed in this study showed a low agreement with each other for estimating cardiovascular risk in the Brazilian population. The highest agreement was observed between the PCE and Globorisk-LAC (AC1 of 0.874). Compared with the Globorisk-LAC, the PCE underestimated the cardiovascular risk in participants with lower risk and overestimated in those with higher risk; the latter caused a change in cardiovascular risk stratification. Moreover, the Framingham overestimated the cardiovascular risk in over 95% of participants compared with the PCE (AC1 of 0.489) and Globorisk-LAC (AC1 of 0.454).
Malta et al.41 analyzed the same dataset and reported a 44% prevalence of high cardiovascular risk based on the 2013 AHA,42 38% using the criteria of the Brazilian Society of Cardiology (SBC), and 19% using the Framingham. In the study, the observed agreement between the SBC and AHA criteria was 43% for high risk and 55% for low risk, whereas an agreement of 51% for high risk and 100% for low risk was found between the SBC and Framingham.
The present study observed a 59% agreement between the Framingham and PCE by simultaneously considering the three risk categories defined by the same cutoff point (i.e., < 10%, 10 to 20%, ≥ 20%), uniformly applied across the scores. Additionally, the analysis used the most recent version of the 2018 PCE.10 The main objective was to compare the risk scores, whereas Malta et al. focused on the limited interchangeability of the classification criteria.
One of the few studies comparing the Globorisk with other scores was conducted by Osei-Yeboah et al.,43 who reported a 13% prevalence of high cardiovascular risk using the Framingham (cutoff point of 20% in 10 years), 0% using the Globorisk (cutoff point of 30%), and 3% using the 2018 PCE (cutoff point of 20%). Cohen's κ coefficient was 0.008 for the agreement between Framingham and Globorisk, 0.432 between the Framingham and PCE, and 0.020 between PCE and Globorisk scores. Another study44 compared the Globorisk (19% high or very high cardiovascular risk) with Framingham (37% high or very high cardiovascular risk) in outpatient clinics for diabetes mellitus in Bangladesh hospitals and observed a Cohen's κ of 0.34.
In our study, the agreement between Framingham and PCE (AC1 = 0.489) was similar to that reported by Osei-Yeboah et al.,43 whereas the agreements between the Globorisk-LAC and Framingham (AC1 = 0.454) and PCE (AC1 = 0.874) were considerably higher. This discrepancy is partly explained by the higher cutoff point (30%) used by Osei-Yeboah et al.43 to define high cardiovascular risk according to the Globorisk, which led to none of the nearly 14,000 participants being classified as high risk. In contrast, the other study44 used the same cutoff point and found an agreement between Globorisk and Framingham similar to that observed in our study.
Notably, both studies43,44 used the global version of the Globorisk13,14 derived primarily from cohorts conducted in high-income countries. To our knowledge, the present study was the first to estimate agreements between other risk scores and the Globorisk-LAC,15,17 which was specifically developed using cohorts from Latin America and the Caribbean.
Several studies have compared the Framingham with the PCE for cardiovascular risk stratification. In studies conducted in outpatient clinics,45-48 the prevalence of high cardiovascular risk ranged from 2% to 50% using the Framingham and from 5% to 50% using PCE, with Cohen's κ values ranging from 0.049 to 0.745. In contrast, population-based studies43,49-51 reported a high-risk prevalence ranging from 3% to 19% using the Framingham and less than 1% to 10% using the PCE (Cohen's κ between 0.29 and 0.55).
A low Cohen's κ was observed in clinical and population-based studies with a lower prevalence of high cardiovascular risk. This illustrates the Cohen's κ bias, which underestimates the agreement when one of the categories has a low prevalence.27,28 This is why our study opted for Gwet's AC127 as the estimand for the agreement in cardiovascular risk stratification.
This study was the first to compare the Framingham, PCE, and Globorisk-LAC scores using the Bland & Altmann method.24 At least one previous study used this method to compare the Globorisk and WHO scores.52 Nevertheless, our study used the ratio rather than the difference between risk estimates and compared it with a predefined range (0.8 to 1.25) to estimate the percentage of cases in which the scores agreed with each other.
Some limitations must be acknowledged. The 2013 PNS laboratory testing database lacked information on clusters, precluding statistical inferences involving confidence intervals or p-values. Nevertheless, this database is the most recent population-representative survey in Brazil that included the necessary laboratory data to estimate cardiovascular risk. Although the low response rate of the laboratory component may have introduced bias, reweighting data likely mitigated this effect. It should also be emphasized that the present study neither verified the predicted outcomes nor determined which scores had superior calibration and discrimination. Instead, the study was limited to evaluating the agreement between cardiovascular risk scores.
To avoid excessive comparisons, the analysis was limited to three cardiovascular risk scores, excluding the WHO score and the recent PREVENT tool. As the WHO score12 is part of the HEARTS strategy53 and demonstrated superior performance in the study by Camargos et al.,16 more studies are needed with the Brazilian population. Recalibration of the PREVENT tool, which incorporated innovative concepts54 and showed exemplary performance,55 would also be required before it could be considered for use in Brazil. This effort may be worthwhile, as the new tool disagreed with the PCE in estimating the cardiovascular risk of millions of people in the United States.56
The substantial disagreement among cardiovascular risk scores when applied to the Brazilian population may be justified by two main factors. First, each score was derived from different cohorts and calibrated for distinct populations. This explains, for example, why the PCE is now used in the United States instead of the Framingham42 and why our study evaluated the consequences of replacing the Framingham with Globorisk-LAC in Brazilian guidelines for dyslipidemia.4,8
Second, the scores differ in the composition of predicted cardiovascular events. The PCE and the Globorisk-LAC estimate the risk of hard atherosclerotic events: nonfatal myocardial infarction, coronary death, and ischemic or hemorrhagic stroke.10,15,42 In contrast, the Framingham also includes coronary insufficiency, angina, transient ischemic attack, peripheral arterial disease, and heart failure.9
These two factors have important consequences for a potential update of clinical practice guidelines,4,8 particularly regarding adapting international recommendations to the Brazilian context. For example, based on the PCE, the pharmacological prevention of atherosclerotic disease in the United States is currently recommended at thresholds of 7.5%5 or 10%26 of 10-year cardiovascular risk. Regarding the interpretation of cardiovascular risk scores, the same cutoff points could be used in Brazil if the risk is estimated using the Globorisk-LAC. On the other hand, this could not be applied to the Framingham, which does not have a version specifically designed for atherosclerotic risk9 and has never been recalibrated for the Brazilian population.11
Conclusion
In the Brazilian population, the estimated 10-year cardiovascular risk substantially differed among the Framingham, PCE, and Globorisk-LAC scores. Given its derivation from Latin American cohorts and its similarity to PCE estimates, the Globorisk-LAC is an excellent option to succeed the Framingham in future updates of the Brazilian Guidelines for Dyslipidemia and Atherosclerosis Prevention, including the associated Clinical Protocols and Therapeutic Guidelines.
The content is available at the DOI https://doi.org/10.5281/zenodo.13948103.
There were no external funding sources for this study.
This study is not associated with any thesis or dissertation work.
This study was approved by the Comissão Nacional de Ética em Pesquisa under the protocol number 328,159. All the procedures in this study were in accordance with the 1975 Helsinki Declaration, updated in 2013. Informed consent was obtained from all participants included in the study.
The authors did not use any artificial intelligence tools in the development of this work.