PEDIATRICS Vol. 106 No. 4 October 2000, pp. 863

Timing of Brain Injury

Hayakawa et al¹ present data in Table 2 of their article describing the test characteristics associated with their proposed method for determining the timing of brain injury in some preterm infants. The following tables present a more explicit analysis of the same (aggregated) data, facilitating computation of a posttest probability.

ECG Result "Gold Standard" Test Results Row Totals PVL or CP Present No PVL or CP Abnormal 29 12 41 Normal 5 118 123 Column totals 34 130 164 PVL indicates periventricular leukomalacia; CP, cerebral palsy.

At least for their study population, the probability that an infant with an abnormal ECG (operationally defined by the authors) has PVL or CP equals:

Using the same method, it is informative to compute the predictive value of their ECG assessment, stratified by ECG findings (ie, acute stage, chronic stage, and both). Thus, the probability that an infant with only acute changes has PVL or CP is:

ECG Result "Gold Standard" Test Results Row Totals PVL or CP Present No PVL or CP Acute stage only 7 7 14 All other results 27 123 150 Column totals 34 130 164

The probability that an infant with chronic stage only has PVL or CP is:

ECG Result "Gold Standard" Test Results Row Totals PVL or CP Present No PVL or CP Chronic stage only 6 3 9 All other results 28 127 155 Column totals 34 130 164

The probability that an infant with both acute and chronic stages has PVL or CP is:

ECG Result "Gold Standard" Test Results Row Totals PVL or CP Present No PVL or CP Acute and chronic stage 16 2 18 All other results 18 128 146 Column totals 34 130 164

This information may further clarify the clinical applicability of the authors' findings.

Joseph Schulman
Albany Medical College
Department of Pediatrics
Division of Neonatal/Perinatal Medicine
Albany, NY 12208-3479

REFERENCE

1. Hayakawa F, Okumura A, Kato T, Kuno K, Watanabe K Determination of timing of brain injury in preterm infants with periventricular leukomalacia with serial neonatal electroencephalography. Pediatrics. 1999; 104:1077-1081 [Abstract/Free Full Text]

In Reply.

Dr Schulman has demonstrated in the letter that the probability that an infant with both acute and chronic stage EEG abnormalities will develop periventricular leukomalacia (PVL) or cerebral palsy (CP) is 0.89. This value is higher than the probability that an infant with acute and/or chronic stage EEG abnormalities will develop PVL or CP. Dr Schulman suggested that this approach might enhance the usefulness of EEG. We agree with Dr Schulman's opinion in this view. The main purpose of our study was to presume whether the brain insult responsible for PVL occurred antenatally, perinatally, or postnatally. Therefore, we did not investigate the predictive value of EEG assessment in detail in this study. This will be the subject of another study.

However, there is a weak point in this approach. The sensitivity of EEG is decreased from 0.85 (29 of 34) in the original data to 0.47 (16 of 34) in the modified data. We consider that sensitivity is one of the most important advantages of EEG.¹ Our unpublished preliminary data revealed that a false-negative rate of ultrasonography for later CP was approximately 0.3, whereas that of EEG was below 0.1. There are some reasons for the decrease in the sensitivity of EEG in the modified approach. First, acute stage EEG abnormalities are not always present in infants with antenatal injury. Severe PVL can later develop in such infants. Second, infants with mild PVL sometimes had chronic stage EEG abnormalities but not acute stage EEG abnormalities. We speculate that acute stage EEG abnormalities may be too subtle to be detected. Infants with both acute and chronic stage EEG abnormalities are likely to represent those with relatively severe PVL whose timing of injury is presumed perinatal. The severity of acute or chronic stage EEG abnormalities is varied among infants. When only severe abnormalities are judged as positive, the specificity of EEG for the later development of PVL is increased, but the sensitivity decreased. In our study, we accentuated the usefulness of EEG for determining the timing of injury, because this information is not obtained from neuroimaging or laboratory studies. However, as Dr Schulman pointed out, the accuracy of EEG in predicting neurodevelopmental outcome is to be investigated.

Akihisa Okumura
Department of Pediatrics
Nagoya University School of Medicine
Nagoya, Aichi, Japan

Fumio Hayakawa
Department of Pediatrics
Okazaki City Hospital
Okazaki, Aichi, Japan

REFERENCE

1. Watanabe K, Hayakawa F, Okumura A Neonatal EEG: a powerful tool in the assessment of brain damage in preterm infants. Brain Dev. 1999; 21:361-372 [CrossRef][Medline]