Abstract:

The increase in the number of reported cases of microcephaly in Pernambuco State, and Northeast Brazil, characterized an epidemic that led the Brazilian Ministry of Health to declare a national public health emergency. The Brazilian Ministry of Health initially defined suspected cases as newborns with gestational age (GA) ≥ 37 weeks and head circumference (HC) ≤ 33cm, but in December 2015 this cutoff was lowered to 32cm. The current study aimed to estimate the accuracy, sensitivity, and specificity of different cutoff points for HC, using ROC curves, with the Fenton and Intergrowth (2014) curves as the gold standard. The study described cases reported in Pernambuco from August 8 to November 28, 2015, according to sex and GA categories. The Fenton and Intergrowth methods provide HC growth curves according to GA and sex, and microcephaly is defined as a newborn with HC below the 3^rd percentile in these distributions. Of the 684 reported cases, 599 were term or post-term neonates. For these, the analyses with ROC curves show that according to the Fenton criterion the cutoff point with the largest area under the ROC curve, with sensitivity greater than specificity, is 32cm for both sexes. Using the Intergrowth method and following the same criteria, the cutoff points are 32cm and 31.5cm for males and females, respectively. The cutoff point identified by the Fenton method (32cm) coincided with the Brazilian Ministry of Health recommendation. Adopting Intergrowth as the standard, the choice would be 32cm for males and 31.5cm for females. The study identified the need to conduct critical and on-going analyses to evaluate cutoff points, including other characteristics for microcephaly case definition.

Keywords:
Microcephaly; Zika Virus Infection; Epidemiological Surveillance; Data Accuracy

Introduction

The rapid rise in the number of suspected cases of microcephaly in live newborns starting in August 2015 in Pernambuco State, Northeast Brazil, called the attention of physicians in the public and private health care systems in that state ¹1. Sene A. Ministério da Saúde investiga aumento de casos de microcefalia em Pernambuco. Caderno Vida Urbana. Diário de Pernambuco 2015. http://www.diariodepernambuco.com.br/app/noticia/vida-urbana/2015/11/04/interna_vidaurbana,608239/ministerio-da-saude-investiga-aumento-de-casos-de-microcefalia-em-pernambuco.shtml (acessado em 04/Nov/2015).
http://www.diariodepernambuco.com.br/app... .

Alerted to the problem, the Executive Secretariat for Health Surveillance (SEVS) of the Pernambuco State Health Secretariat (SES-PE) detected a change in the pattern of occurrence of this congenital anomaly, with an increase in the number of cases when compared to previous years, thus characterizing an epidemic ²2. Secretaria Executiva de Vigilância em Saúde, Secretaria Estadual de Saúde. Possível alteração do padrão de ocorrência de microcefalia em nascidos vivos no Estado de Pernambuco. Recife: Secretaria Estadual de Saúde; 2015. (Nota Técnica, 43/15)..

The possible association between the increase in the number of microcephaly cases and outbreaks of Zika virus infection in Brazil since late 2014, especially in Northeast Brazil, became the object of investigation ³3. Ministério da Saúde. Ministério da Saúde confirma relação entre vírus Zika e microcefalia. http://portalsaude.saude.gov.br/index.php/o-ministerio/principal/secretarias/svs/noticias-svs/21016-ministerio-da-saude-confirma-relacao-entre-virus-zika-e-microcefalia (acessado em 30/Nov/2015).
http://portalsaude.saude.gov.br/index.ph... .

Other states of Northeast Brazil reported an increase in microcephaly cases, leading the Brazilian Ministry of Health to declare a national public health emergency on November 11, 2015 ⁴4. Secretaria de Vigilância em Saúde, Ministério da Saúde. Procedimentos preliminares a serem adotados para a vigilância dos casos de microcefalia no Brasil. Brasília: Ministério da Saúde; 2015. (Nota Informativa, 01/2015 - COES Microcefalias)..

For reporting purposes, suspected microcephaly cases were initially defined as live newborns with gestational age ≥ 37 weeks and head circumference (HC) ≤ 33cm ⁵5. Secretaria Executiva de Vigilância em Saúde, Secretaria Estadual de Saúde. Protocolo clínico e epidemiológico para investigação de casos de microcefalia. Versão no 1. Recife: Secretaria Estadual de Saúde; 2015.. However, this cutoff point, with high sensitivity but not fully backed by the existing scientific literature ⁶6. Fenton TR, Kim JH. A systematic review and meta-analysis to revise the Fenton growth chart for preterm infants. BMC Pediatrics 2013; 13:59.^,77. Villar J, Ismail LC, Victora CG, Ohuma EO, Bertino E, Altman DG, et al. International standards for newborn weight, length, and head circumference by gestational age and sex: the Newborn Cross-Sectional Study of the INTERGROWTH-21st Project. Lancet 2014; 384:857-68., generated an excessive number of notifications. Thus, as of December 2015, the Brazilian Ministry of Health established HC ≤ 32cm as the definition of a suspected case of microcephaly ⁸8. Secretaria Executiva de Vigilância em Saúde, Secretaria Estadual de Saúde. Protocolo clínico e epidemiológico para investigação de casos de microcefalia. Versão no 2. Recife: Secretaria Estadual de Saúde; 2015..

To contribute to the discussion on the cutoff point for reporting suspected cases, the current study aimed to estimate the accuracy, sensitivity, and specificity of different cutoff points using the Fenton method and the curve proposed by the Intergrowth project ⁷7. Villar J, Ismail LC, Victora CG, Ohuma EO, Bertino E, Altman DG, et al. International standards for newborn weight, length, and head circumference by gestational age and sex: the Newborn Cross-Sectional Study of the INTERGROWTH-21st Project. Lancet 2014; 384:857-68. as the gold standard. The Intergrowth curve was included in the analyses because it was developed recently, based on measurements of children from countries with different ethnic and economic characteristics.

Material and methods

The study analyzed a total of 696 suspected microcephaly cases reported to the SES-PE from August 2 to November 28, 2015 (epidemiological weeks 31 to 47).

The analyses excluded newborns with no recorded information on head circumference and/or gestational age and another two cases classified as recording errors (HC = 35cm and 45cm), leaving two cases with HC = 33.5cm, identified in the sample as suspected cases of microcephaly, leaving a total of 684 suspected cases.

The study began with a description of reported cases by distributions according to sex and gestational age categories.

Gestational age (GA) categories were defined as follows: preterm neonates - GA up to 36 weeks and 6 days; term and post-term neonates - GA 37 weeks or more, in which term neonates were defined as GA up to 41 weeks and 6 days.

To classify suspected microcephaly cases as positive or negative, we first used the Fenton tables and then the Intergrowth method (2014), which provide distributions of growth curves for head circumference according to GA and sex; positive (confirmed) cases were defined as newborns with HC below the 3rd percentile of these distributions. We also analyzed the time trend in suspected and confirmed cases according to epidemiological week.

ROC (Receiver Operating Characteristic) curves were used to evaluate the cutoff points for HC. The gold standard was first the Fenton curves and then the Intergrowth tables.

The study also analyzed differences in the proportion of confirmed cases according to the Fenton curve, by sex.

The criterion for selecting the "ideal" cutoff point was maximization of the area under the ROC curve, respecting the need for greater sensitivity than specificity, given the nature of the object of investigation (screening). The analyses were performed separately according to sex and GA categories.

The Intergrowth method, at the beginning of the epidemic, only provided growth curves of HC for new borns with GA ≥ 33 weeks. For consistency, only neonates with GA ≥ 37 weeks were included in the analyses using this method.

Finally, the study verified the concordance between the classifications obtained by the two methods, using the kappa coefficient. The analyses were performed with Stata, version 12 (StataCorp LP, College Station, USA).

Results

Of the 684 reported cases, 599 were term/post-term neonates and 85 (12.4%) were preterm, a similar percentage to that of the general population according to the Brazilian Information System on Live Births (SINASC) for Pernambuco State in 2013. Seven preterm neonates had GA between 22 and 31 weeks. The majority of the reported cases were females (423 cases, or 62% of the total).

Classification of reported cases according to the Fenton criterion by sex and GA showed a positive rate of 39% (267 cases), leaving 417 false-positives (61%). For the 599 neonates with GA ≥ 37 weeks, the positive rate according to the Fenton criterion was 41% (243 cases), while the Intergrowth method showed 188 positive cases (31%).

The percentages of positive microcephaly cases did differ significantly according to sex: 41% for males and 38% for females (χ²(1 df) = 0.44; p = 0.506).

Notwithstanding the high number of false-positives, the 267 confirmed cases according to the Fenton criterion, reported over the course of 17 epidemiological weeks, represent a relative rate of 58 cases per 10,000 live births.

Figure 1 shows the distribution of suspected and confirmed cases according to the Fenton method, by epidemiological week.

Table 1 shows the sensitivity, specificity, and accuracy using ROC curves with the two methods. According to the Fenton criterion, the cutoff point with the largest area under the ROC curve, and respecting the need for higher sensitivity than specificity, was 32cm for both sexes. However, for females this cutoff point showed a specificity of 70%, while for males it was approximately 80%.

According to the Intergrowth method, the cutoff points using the same criteria are 32cm and 31.5cm for males and females, respectively.

Concordance between the methods for both sexes was 90.8% (kappa coefficient = 0.802; p < 0.001). Concordance was 90.3% for males (kappa = 0.792; p < 0.001) and 91.2% for females (kappa = 0.809; p < 0.001).

Table 2 shows the classification of newborns according to cutoff points for HC identified by the two methods in the analysis of accuracy according to GA categories. For newborns with GA ≥ 37 weeks, the Intergrowth method shows a false-positive rate of 11.4% (68 cases), compared to 15.7% by the Fenton method. Considering newborns of all gestational ages, there were a total of 108 false-positives (15.8%) according to the Fenton method.

Finally, considering newborns with GA ≥ 37 weeks (599), a cutoff of 32cm would lead to 54% positives (324 cases), while cutoffs with higher specificity, like 31cm and 30.5cm, would lead to 29% and 20% positives, respectively.

Discussion

Using the Brazilian Ministry of Health criterion (HC ≤ 33cm), classification of the Fenton curves for the 684 newborns reported as suspected cases of microcephaly shows 417 false-positives (61%).

Meanwhile, analysis of the ROC curves using Fenton as the gold standard points to a cutoff point of 32cm for HC in both sexes, consistent with the new recommendation by the Ministry of Health ⁸8. Secretaria Executiva de Vigilância em Saúde, Secretaria Estadual de Saúde. Protocolo clínico e epidemiológico para investigação de casos de microcefalia. Versão no 2. Recife: Secretaria Estadual de Saúde; 2015..

Lowering the cutoff point for HC from 33cm to 32cm increased the accuracy of the case definition for microcephaly, by sex and GA, according to the Fenton method. Among term/post-term neonates of both sexes, the number of false-positives fell from 275 (45.9%) to 94 (15.7%).

Still, the adoption of a single cutoff point (32cm) for both sexes does not appear appropriate ⁹9. Woods CG, Parker A. Investigating microcephaly. Arch Dis Child 2013; 98:707-13., since it results in a 10% drop in specificity for female term and post-term neonates, producing more false-positives in the latter group. This finding explains the large volume of reported suspected cases of microcephaly in females, despite the lack of a significant difference between the sexes in the proportion of positive cases.

Considering preterm neonates, a cutoff of 30cm produced 14 false-positives among the 85 suspected cases.

Since it shows higher specificity, the Intergrowth method pointed to HC = 31.5cm as the cutoff that produces the largest area under the ROC curve for males, compared to 31cm for females. However, since screening requires greater sensitivity than specificity, the choice would be 32cm for males and 31.5cm for females. Importantly, variation in HC measurements is possible in this epidemic, due to phenotypical characteristics such as excess scalp skin as a result of cerebral atrophy ¹⁰10. Schuler-Faccini L, Ribeiro EM, Feitosa IML, Horovitz DDG, Cavalcanti DP, Pessoa A, et al. Possible association between Zika virus infection and microcephaly - Brazil, 2015. MMWR Morb Mortal Wkly Rep 2016; 65:59-62.. The cutoff for HC in newborns should be sensitive enough to generate an excessive number of notifications. Adoption of the third percentile for the curves from the two methods as the classification criterion would mean reporting 3% of the term birth cohort as suspected cases of microcephaly ¹¹11. Victora CG, Schuler-Faccini L, Matijasevich A, Ribeiro E, Pessoa A, Barros FC. Microcephaly in Brazil: how to interpret reported numbers? Lancet 2016; 387:621-4.. Meanwhile, the recommendation of lower cutoff points, that is, more specific and which would include lower percentages of births for investigation, might exclude cases that would warrant investigation to identify possible abnormalities.

In conclusion, critical and on-going analysis is necessary to support surveillance of microcephaly cases in newborns in the context of the current epidemic. Other clinical and/or phenotypical criteria should be explored, as well as findings from imaging tests, in order to establish a more accurate gold standard and improve the reporting, investigation, and treatment of cases. These aspects are crucial for improving knowledge of this new syndrome, probably associated with the Zika virus infection.

Acknowledgments

The authors wish to acknowledge the support of the Brazilian Ministry of Health, the Pernambuco State Health Secretariat (SES-PE), and the Pan American Health Organization. The authors received partial support from the National Council for Scientific and Technological Development - CNPq (grants 306489/2010-4 for C.M.T.M., 308311/2009-4 for R.A.A.X., 308590/2013-9 for D.B.M.F., 301779/2009-0 for M.F.P.M.A., 304174/2014-9 for M.C.B., and 306222/2013-2 for W.V.S.). We also wish to thank Dr. Enrique Vázquez (Pan American Health Organization, Brasília, Brazil) for his attentive reading of the manuscript and important suggestions

References

Publication Dates

History