Vania Naomi Hirakata

Bias of using odds ratio estimates in multinomial logistic regressions to estimate relative risk or prevalence ratio and alternatives

Recent studies have emphasized that there is no justification for using the odds ratio (OR) as an approximation of the relative risk (RR) or prevalence ratio (PR). Erroneous interpretations of the OR as RR or PR must be avoided, as several studies have shown that the OR is not a good approximation for these measures when the outcome is common (> 10%). For multinomial outcomes it is usual to use the multinomial logistic regression. In this context, there are no studies showing the impact of the approximation of the OR in the estimates of RR or PR. This study aimed to present and discuss alternative methods to multinomial logistic regression based upon robust Poisson regression and the log-binomial model. The approaches were compared by simulating various possible scenarios. The results showed that the proposed models have more precise and accurate estimates for the RR or PR than the multinomial logistic regression, as in the case of the binary outcome. Thus also for multinomial outcomes the OR must not be used as an approximation of the RR or PR, since this may lead to incorrect conclusions.Recentes trabalhos tem enfatizado que ja nao ha justificativa para o uso da razao de chances (RC) como aproximacao do risco relativo (RR) ou razao de prevalencia (RP). Deve-se evitar a interpretacao equivocada da RC como RR ou RP, pois varios estudos demonstraram que a RC nao e uma boa aproximacao para tais medidas quando o desfecho e comum (> 10%). Para desfechos multinomiais e usual aplicar a regressao logistica multinomial. Nesse contexto, nao ha estudos demonstrando o impacto da aproximacao da RC nas estimativas de RR ou RP. O objetivo deste trabalho e apresentar e discutir metodos alternativos a regressao logistica multinomial, baseados na regressao de Poisson e no modelo log-binomial. As abordagens foram comparadas por um estudo de simulacao com diversos cenarios. Assim como no caso do desfecho binario, os modelos propostos apresentaram estimativas mais precisas e acuradas para o RR ou RP do que a regressao logistica multinomial. Entao, tambem para os desfechos multinomiais nao se deve utilizar a RC como aproximacao do RR ou RP, pois conclusoes incorretas podem ocorrer.

Clinical characteristics of women with gestational diabetes - comparison of two cohorts enrolled 20 years apart in southern Brazil

CONTEXT AND OBJECTIVE:nThe prevalence and characteristics of gestational diabetes mellitus (GDM) have changed over time, reflecting the nutritional transition and changes in diagnostic criteria. We aimed to evaluate characteristics of women with GDM over a 20-year interval.nnnDESIGN AND SETTING:nComparison of two pregnancy cohorts enrolled in different periods, in university hospitals in Porto Alegre, Brazil: 1991 to 1993 (n = 216); and 2009 to 2013 (n = 375).nnnMETHODS:nWe applied two diagnostic criteria to the cohorts: International Association of Diabetes and Pregnancy Study Groups (IADPSG)/World Health Organization (WHO); and National Institute for Health and Care Excellence (NICE). We compared maternal-fetal characteristics and outcomes between the cohorts and within each cohort.nnnRESULTS:nThe women in the 2010s cohort were older (31 ± 7 versus 30 ± 6 years), more frequently obese (29.4% versus 15.2%), with more hypertensive disorders (14.1% versus 5.6%) and at increased risk of cesarean section (adjusted relative risk 1.8; 95% confidence interval: 1.4 - 2.3), compared with those in the 1990s cohort. Neonatal outcomes such as birth weight category and hypoglycemia were similar. In the 1990s cohort, women only fulfilling IADPSG/WHO or only fulfilling NICE criteria had similar characteristics and outcomes; in the 2010s cohort, women only diagnosed through IADPSG/WHO were more frequently obese than those diagnosed only through NICE (33 ± 8 kg/m2 versus 28 ± 6 kg/m2; P < 0.001).nnnCONCLUSION:nThe epidemic of obesity seems to have modified the profile of women with GDM. Despite similar neonatal outcomes, there were differences in the intensity of treatment over time. The IADPSG/WHO criteria seemed to identify a profile more associated with obesity.

Birth weight classification in gestational diabetes: is there an ideal chart?

Introduction : Gestational diabetes mellitus (GDM) is associated to increased rates of large for gestational age newborns and macrosomia. Several charts are used to classify birth weight. Is there an ideal chart to classify newborns of GDM mothers? Methods : We evaluated adequacy of birth weight of 332 neonates born to GDM mothers at Hospital de Clinicas de Porto Alegre, Brazil. Newborns were classified according to gestational age as small (SGA), adequate or large (LGA) based on four charts: Alexander, Pedreira, INTERGROWTH 21 st Project and SINASC-2012. The latter was built using data from a large national registry of 2012, the Born Alive National Surveillance System ( Sistema de Informacoes de Nascidos Vivos – SINASC), which included 2.905,789 birth certificates. Frequencies of SGA and LGA and Kappa agreement were calculated. Results : In non-gender adjusted curves, SGA rates (95% confidence interval) varied from 8% (5-11) to 9% (6-13); LGA rates, from 11% (8-15) to 17% (13-21). For males, SGA rates varied from 3% (1-6%) to 6% (3-11%), and LGA rates, from 18% (13-24%) to 31% (24-38%); for female, SGA rates were from 3% (1-7%) to 10% (6-16%) and LGA rates, from 11% (6-16%) to 19% (13-26%). Kappa results were: ALEXANDER vs. SINASC-2012: 0.80 (0.73-0.88); INTERGROWTH 21 st vs. SINASC-2012 (adjusted by sex): 0.62 (0.53-0.71); INTERGROWTH 21 st vs. PEDREIRA: 0.71 (0.62-0.79); SINASC-2012 (by sex) vs. PEDREIRA: 0.86 (0.79-0.93). Conclusions : Misclassification has to be taken into account when evaluating newborns of GDM mothers, as LGA rates can almost double depending on the chart used to classify birth weight.

Vies da razao de chances estimada pela regressao logistica multinomial para estimar o risco relativo ou a razao de prevalencia e alternativas

Recent studies have emphasized that there is no justification for using the odds ratio (OR) as an approximation of the relative risk (RR) or prevalence ratio (PR). Erroneous interpretations of the OR as RR or PR must be avoided, as several studies have shown that the OR is not a good approximation for these measures when the outcome is common (> 10%). For multinomial outcomes it is usual to use the multinomial logistic regression. In this context, there are no studies showing the impact of the approximation of the OR in the estimates of RR or PR. This study aimed to present and discuss alternative methods to multinomial logistic regression based upon robust Poisson regression and the log-binomial model. The approaches were compared by simulating various possible scenarios. The results showed that the proposed models have more precise and accurate estimates for the RR or PR than the multinomial logistic regression, as in the case of the binary outcome. Thus also for multinomial outcomes the OR must not be used as an approximation of the RR or PR, since this may lead to incorrect conclusions.Recentes trabalhos tem enfatizado que ja nao ha justificativa para o uso da razao de chances (RC) como aproximacao do risco relativo (RR) ou razao de prevalencia (RP). Deve-se evitar a interpretacao equivocada da RC como RR ou RP, pois varios estudos demonstraram que a RC nao e uma boa aproximacao para tais medidas quando o desfecho e comum (> 10%). Para desfechos multinomiais e usual aplicar a regressao logistica multinomial. Nesse contexto, nao ha estudos demonstrando o impacto da aproximacao da RC nas estimativas de RR ou RP. O objetivo deste trabalho e apresentar e discutir metodos alternativos a regressao logistica multinomial, baseados na regressao de Poisson e no modelo log-binomial. As abordagens foram comparadas por um estudo de simulacao com diversos cenarios. Assim como no caso do desfecho binario, os modelos propostos apresentaram estimativas mais precisas e acuradas para o RR ou RP do que a regressao logistica multinomial. Entao, tambem para os desfechos multinomiais nao se deve utilizar a RC como aproximacao do RR ou RP, pois conclusoes incorretas podem ocorrer.

El sesgo del odds ratio estimado por la regresión logística multinomial para estimar el riesgo relativo o la razón de prevalencia y alternativas

Recent studies have emphasized that there is no justification for using the odds ratio (OR) as an approximation of the relative risk (RR) or prevalence ratio (PR). Erroneous interpretations of the OR as RR or PR must be avoided, as several studies have shown that the OR is not a good approximation for these measures when the outcome is common (> 10%). For multinomial outcomes it is usual to use the multinomial logistic regression. In this context, there are no studies showing the impact of the approximation of the OR in the estimates of RR or PR. This study aimed to present and discuss alternative methods to multinomial logistic regression based upon robust Poisson regression and the log-binomial model. The approaches were compared by simulating various possible scenarios. The results showed that the proposed models have more precise and accurate estimates for the RR or PR than the multinomial logistic regression, as in the case of the binary outcome. Thus also for multinomial outcomes the OR must not be used as an approximation of the RR or PR, since this may lead to incorrect conclusions.Recentes trabalhos tem enfatizado que ja nao ha justificativa para o uso da razao de chances (RC) como aproximacao do risco relativo (RR) ou razao de prevalencia (RP). Deve-se evitar a interpretacao equivocada da RC como RR ou RP, pois varios estudos demonstraram que a RC nao e uma boa aproximacao para tais medidas quando o desfecho e comum (> 10%). Para desfechos multinomiais e usual aplicar a regressao logistica multinomial. Nesse contexto, nao ha estudos demonstrando o impacto da aproximacao da RC nas estimativas de RR ou RP. O objetivo deste trabalho e apresentar e discutir metodos alternativos a regressao logistica multinomial, baseados na regressao de Poisson e no modelo log-binomial. As abordagens foram comparadas por um estudo de simulacao com diversos cenarios. Assim como no caso do desfecho binario, os modelos propostos apresentaram estimativas mais precisas e acuradas para o RR ou RP do que a regressao logistica multinomial. Entao, tambem para os desfechos multinomiais nao se deve utilizar a RC como aproximacao do RR ou RP, pois conclusoes incorretas podem ocorrer.