Volume 83, Issue 1, Pages 63-68 (January 2007)

12 of 12

ABSTRACT

FULL TEXT

FULL-TEXT PDF (144 KB)

Isolated parenchymal lesions on cranial ultrasound in very preterm infants in the context of maternal infection

Abstract

Aim

To explore the associations between a clinical diagnosis of maternal infection (CDMI) and findings on the initial cranial ultrasound scan in very preterm infants.

Methods

Among infants born at less than 32 weeks gestation, cases of CDMI and controls were identified on the basis of routinely available obstetric data. Neonatal cranial ultrasound scans carried out soon after birth were retrospectively reviewed for evidence of parenchymal echodensity (PED), intraventricular haemorrhage (IVH) or PED contiguous with IVH.

Results

Any PED was identified in 20/40 (50%) cases of CDMI and 9/30 (30%) of controls. Logistic regression was used to adjust for differences between the two study groups. When compared with normal scans, isolated PED was more likely with CDMI odds ratio, OR (95% confidence interval, CI), 41.8 (2.64, 662) and lower Apgar score at 5 min 2.89 (1.05, 7.98). IVH was more likely with lower gestational age, OR for each completed week of gestation 0.64 (0.46, 0.88) and a protective effect of female sex, OR 0.25 (0.063, 0.98), PED contiguous with IVH was more likely with lower gestational age OR 0.59 (0.336, 1.04).

Conclusions

CDMI may be associated with isolated PED in very preterm infants. We speculate that isolated PED (including “flares”) identify infants who have sustained early brain injury because of intrauterine infection. Isolated PED may be a useful intermediate outcome in perinatal cohort studies.

Article Outline

• Abstract

• 1. Methods

• 1.1. Maternal infection

• 1.2. Neonatal condition

• 1.3. Cranial ultrasound assessment

• 1.4. Plan of analysis

• 2. Results

• 3. Discussion

• Acknowledgment

• References

• Copyright

The association between intra-uterine infection and neonatal brain injury was strengthened by a report about findings on magnetic resonance imaging (MRI) performed soon after birth. Specifically, evidence of antenatal immune activation was associated with abnormalities in the postnatal appearances of brain parenchyma which could occur with, or without, intraventricular haemorrhage (IVH) or cystic change [1]. MRI may be a useful modality to demonstrate the consequences of antenatal inflammation but is more difficult to perform than the cranial ultrasound scans routinely available in neonatal units. Findings on cranial ultrasound (such as non-cystic changes in the brain parenchyma not in continuity with IVH) are likely to be analogous to parenchymal abnormalities on MRI. Non-cystic parenchymal findings on cranial ultrasound have been given a variety of labels such as hyperechoic flares or transient hyperechoic lesions [2]. It has been suggested that the term “parenchymal echodensities” (PED) may be the most appropriate way to describe this group of sonographic appearances since the term “PED” does not imply any particular underlying mechanism or location [3], [4]. It is currently unclear whether isolated PED (be they transient “flares” or longer-lasting lesions) are related to antenatal circumstances [2]. Given that cranial ultrasound would be much better suited to large-scale studies of perinatal brain injury than MRI, we examined the hypothesis that isolated PED, or other ultrasound findings, are more likely than normal scans in the presence of clinical evidence of intra-uterine infection.

A variety of markers for intra-uterine infection have been used in previous studies. In this exploratory study, we elected to use a clinical diagnosis of maternal infection (CDMI) as the marker of intra-uterine infection. Using data prospectively recorded in the course of routine clinical practice, we defined a group of infants with CDMI and a group of infants without CDMI or any other evidence of maternal or fetal compromise. In order to avoid confounding from postnatal events, the study was limited to examining the relationships between clinical factors before delivery, clinical factors up to 5 min after delivery (summarised by the Apgar score at 5 min) and the initial routine cranial ultrasound scan.

1. Methods

We defined cases of CDMI and controls on the basis of data prospectively recorded in the course of routine clinical practice (Table 1) and examined the incidence of specific outcomes differed between these two groups. We anticipated that the group of infants with CDMI would not be the same as the group of control infants. Thus, we gathered obstetric and neonatal data that would allow us to control for the differences between the two study groups prior to and immediately after delivery. We collated data concerning the initial cranial ultrasound scan performed on each infant.

Table 1.

Obstetric variables contributing to a clinical diagnosis of maternal infection


	Variable	N (%) in 40 cases of antenatal maternal infection
	Maternal pyrexia (>38 °C)	9 (22)
	Maternal CRP>10	17 (43)
	Positive maternal urine culture	7 (17)
	Positive HVS	27 (68)
	Positive maternal blood culture	2 (5.0)
	“Clinical chorioamnionitis” (foul-smelling liquor, uterine tenderness)	13 (32)
	Any maternal antibiotics	31 (77)
		Median (minimum, maximum) for cases
	Highest antenatal maternal white cell count	18.4 (5.7, 34)

Among mothers of control infants 4/30 (15%) received antepartum antibiotics, for example as part of the ORACLE trial: no mothers of control infants had any of the indicators of infection. Among mothers of control infants the highest antenatal white cell count had a median of 16.4 minimum 7.3, maximum 26.6: the white cell count was lower in control mothers than in mothers of case infants (p<0.05, Mann–Whitney U test).

1.1. Maternal infection

Obstetric notes relating to all admissions to the Neonatal Medical Unit between 1st March 1996 and 30th June 1999 were scrutinised by two experienced obstetricians (SV, EH). In case-notes that were eligible for inclusion in the study, a range of potential markers of maternal infection were recorded using a pre-determined pro-forma (see Table 1). Case definition was based on review of the pro-forma by one experienced obstetrician who decided whether or not the mother clearly had evidence of an infection prior to delivery. Women with infection could have some, or all, of the criteria included on the pro-forma. In some cases, it was not possible to make a clear decision. These cases were labelled as dubious and not included in the analysis. For control infants, no pathology was obvious during the pregnancy. We took care to exclude infants with clear evidence of in utero compromise since previous reports suggested that antenatal conditions other than infection alter the incidence of abnormal findings on cranial ultrasonography: pregnancy induced hypertension and amniotic sac inflammation are associated with a different incidence of IVH [5]. Thus, infants were excluded if antenatal imaging showed growth restriction complicated by abnormal patterns of umbilical artery velocities on Doppler ultrasound. Infants were also excluded if they were twins, triplets or higher order pregnancies, postnatal transfers or antenatal transfers from another unit less than 12 h before delivery. Obstetric data relating to the study groups are shown in Table 2.

Table 2.

Characteristics of cases and controls


Obstetric variables
Variable		N (%) in 40 cases of CDMI	N (%) in 30 controls	P (Fisher's exact test or Chi-squared)	Odds ratio (95% confidence interval)
Any exposure to antenatal steroids		37 (93)	26 (87)	0.45
Any exposure to antenatal NSAID		8 (20)	5 (17)	0.77
Labour	Spontaneous	32 (82)	26 (87)	0.45
	Induced	2 (5)	0 (0)
	Did not labour	5 (13)	4 (13)
Presentation	Cephalic	26 (67)	20 (67)	0.69
	Breech	2 (31)	8 (27)
	Transverse	1 (3)	2 (7)
C. section		7 (18)	9 (30)	0.26
		Median (minimum, maximum, N) for 40 cases	Median (minimum, maximum) for 30 controls	P (Mann–Whitney U test)
Gestational age at delivery		25 (23, 31)	29 (23, 31)	<0.001
Duration of ROM (days)		4 (0, 63)	0 (0, 8)	<0.001
Duration of 2nd stage (min)		9 (0, 34)	6.5 (0, 160)	0.98

Neonatal variables
Variable		N (%) in 40 cases of CDMI	N (%) in 30 controls	P (Fisher's exact test or Chi-squared)	Odds ratio (95% confidence interval)
Small for gestational age (<3rd centile)		4 (11)	0 (0)	0.13
Female		25 (63)	<0.005	<0.005	5.5 (1.9, 15)
		Median (minimum, maximum) for cases	Median (minimum, maximum) for controls	P (Mann–Whitney U test)
Birth weight (kg)		0.92 (0.41, 1.5)	1.3 (0.53, 2.5)	<0.001
Apgar score at 5 min		8 (5, 10)	9 (5, 10)	<0.005

1.2. Neonatal condition

Neonatal notes were scrutinised by an experienced paediatrician (SDS). Neonatal data relating to the study groups are shown in Table 2. The Apgar score at 5 min was used as a summary of the condition of each infant at birth.

1.3. Cranial ultrasound assessment

Cranial ultrasound scans were performed at the earliest opportunity according to the unit policy. An Acuson 128XP machine with a 7.5 MHz probe was used by one of two consultant radiologists with a special interest in neonatal imaging. Ultrasound scans were performed on a median of day 1 in cases of CDMI (minimum day 0, maximum 9.5) and on a median of day 2 in controls (minimum day 0, maximum day 13). The median day of scan was significantly different in cases of CDMI and controls (Mann–Whitney, p<0.05). However, the median day of scan did not differ between scans found to have PED or IVH. The initial cranial ultrasound scan for each infant was reported in retrospect by one experienced neonatal radiologist (SR). Since the focus of this study was on the relationships between antenatal maternal infection and findings on the earliest scan performed as part of routine clinical practice, we recorded the presence or absence of each feature on the initial scan only. Definitions of scan abnormalities were similar to those used in a recent study from this institution [6]. The presence on the initial cranial ultrasound scan of the following were recorded on each side of the brain: IVH—defined as increased echogenicity in the ventricular system separate from the normal choroid plexus; any white matter changes including echolucencies and echodensities. In accord with the suggestions made by Paneth [3], we recorded the presence or absence of echodensity or echolucency in brain tissue without reference to imputed causative processes or other descriptive terms. PED were defined as focal increases in echogenicity in the cerebral cortices (i.e. not in basal ganglia or thalami). Thus, the term “PED” includes what others would describe as “flares” and lesions that others would regard as including calcium. Parenchymal echolucencies were defined as focal decreases in echogenicity. A secondary analysis was based on whether PED was contiguous with IVH. For each infant's initial scan, there were hard copies of at least six planes which would have routinely included any abnormality seen.

1.4. Plan of analysis

Imaging findings for the CDMI group were initially compared to the control group using Fisher's exact test. The differences between the two groups were explored using χ2 or Fisher's exact test for categorical variables and Mann–Whitney tests for ordinal and numerical variables. A multivariate approach was used to examine the associations of our pre-specified outcomes. Since we anticipated imbalance between the two clinical groups, and this imbalance could either hide or accentuate the associations of interest, we examined the associations of all the pre-specified outcomes using logistic regression. Logistic regression was used to examine the associations of specific imaging findings controlling for the differences between cases of CDMI and control infants. In order to minimise the risk of “overfitting” the data, we used a forward logistic regression model and the likelihood ratio to examine whether variables should be included in the model. For each imaging finding infants were categorized into those with that finding and those with a normal scan: those infants with a different imaging finding were not included in that particular model. In order to overcome the effects of imbalance between the groups, variables were available for the regression models if they were associated with CDMI on univariate testing (p<0.1), were relevant to the condition of the infant immediately after delivery and were not relevant to the obstetric decision about the presence or absence of CDMI. Thus, in addition to CDMI, the variables available to each model were: infant sex, duration of ROM, gestational age and Apgar score at 5 min. At each step of forward logistic regression, the entry criterion was p<0.05 and a variable was retained in the model if removing it had a significant effect with p<0.1. Univariate analysis of categorical data was done using StatsDirect 2.3.1. All other analysis was performed using SPSS v10.1.

All assessments were performed independently and blind to the results of the other assessments (the radiologist reported hard copies blind to clinical events). Records were linked by hospital number and date of birth for mother and infant: the study was anonymised. All information was obtained in the course of routine clinical practice. Thus, at the time of the study, ethical approval was not required. Study coordination, data entry and statistical analysis were performed by one investigator who had no part in any of the assessments (MT).

287 infants of less than 32 completed weeks gestation were admitted during the study period (1st March 1996–30th June 1999) to the Neonatal Medical Unit at St. Mary's Hospital Manchester, U.K. During this period, there was a consistent approach to imaging by consultant radiologists. Following exclusion of multiple pregnancies, infants with evidence of maternal or fetal compromise not related to infection and infants who had their initial scan on day 14 or later, 70 infants had hard copies of scans available for report at the time of review.

2. Results

Table 2 gives the clinical characteristics of cases and controls. Of note, infants with CDMI were more likely to be female. PED were found in 20/40 infants with CDMI: in 9/20 PED were isolated, in 11/20 there was also IVH and in 5/11 PED was contiguous with IVH. PED were found in 9/30 infants without CDMI: in 3/9 PED were isolated, in 6/9 there was also IVH and in 3/6 PED was contiguous with IVH. A comparison of these proportions between infants with CDMI and control infants is shown in Table 3. These comparisons indicate that in the absence of adjustment for confounding variables there was a weak association between CDMI and any PED but not for the other ultrasound findings. Four infants had parenchymal echolucencies: all four had no history of CDMI. One infant had IVH in the absence of PED.

Table 3.

The incidence of ultrasound findings in infants with and without CDMI


	N (%) in 40 cases of CDMI	N (%) in 30 controls	P (Fisher's exact test)	Risk ratio (95% confidence interval)
Normal	20/40 (50)	13/30 (43)	0.63	1.1 (0.74, 1.7)
Any PED	20/40 (50)	9/30 (30)	0.09	1.50 (0.98, 2.2)
Isolated PED	9/40 (22)	3/30 (10)	0.22	1.4 (0.84, 2.0)
Any IVH	11/40 (28)	8/30 (27)	0.58	1.0 (0.36, 3.0)
PED+contiguous IVH	5/40 (12)	3/30 (10)	0.99	1.1 (0.53, 1.7)

A series of logistic regression models was used to take account of confounding variables that might be associated with particular scan abnormalities (see Table 4). Each model in Table 4 includes only the infants relevant to a comparison between a normal scan and the scan finding in question so that a different number of infants is included in each model. The results show that any PED was associated with a history of CDMI, albeit with broad confidence intervals. Isolated PED was associated with a history of CDMI and lower Apgar scores at 5 min. IVH was associated with lower gestational age and male sex. PED contiguous with IVH was weakly associated with lower gestational age. The associations of IVH were the same, irrespective of whether infants with subependymal haemorrhage were counted as having IVH (data not shown).

Table 4.

Clinical features associated with abnormal ultrasound findings


Ultrasound finding	Infants included in analysis	Significantly associated variable(s)	P-value for significant variables	Adjusted odds ratio (95% confidence interval)
Any PED	58	Maternal infection	0.022	3.8 (1.2, 12)
Isolated PED	43	Maternal infection	0.008	42 (2.6, 660)
		Apgar score at 5 min	0.012	2.9 (1.0, 8.0)
IVH	58	Gestational age	0.007	0.64 (0.46, 0.88)
		Infant sex	0.047	0.25 (0.063, 0.98)
PED+contiguous IVH	42	Gestational age	0.069	0.59 (0.34, 1.0)

Logistic regression was used to determine the factors predicting the incidence of each scan finding, compared to a normal scan. Thus each model included a different number of infants, as shown in the Table. The variables available to the models were: infant sex, gestational age, duration of PPROM, clinical diagnosis of maternal infection and Apgar score at 5 min.

3. Discussion

Each ultrasound finding was associated with a specific pattern of vulnerability. In particular, PED in the absence of IVH were associated with CDMI. This report has three sources of novelty. Firstly, the study groups differed from previous comparisons which have included infants affected by other sources of pathology relevant to scan findings, e.g. pre-eclampsia [5]. Secondly, previous studies of the antenatal associations of isolated parenchymal lesions have not been limited to premature infants [7], or have not included significant numbers of infants exposed to intrauterine infection [8]. Thirdly, this exploratory study of the relevance of routinely performed ultrasound findings to studies of in utero infection used maternal information routinely recorded in clinical notes. Although other means of identifying perinatal inflammation have been described, each has imperfect predictive power and none are universally available in clinical practice.

This study illustrates the feasibility of developing an interim marker of brain injury following clinically evident maternal infection. This marker (isolated PED) could be used to examine how subsequent events modify the extent of brain injury. By defining the study groups on the basis of prospectively recorded data regarding the exposure of interest and collating each aspect of the data blind to the other aspects, we aimed to avoid many of the biases that could arise from a retrospective study design (e.g. case-control study based on the outcome of interest). The pattern of care could affect the results; however, antenatal corticosteroids and prophylactic surfactant were both established elements of care during the time period of this study. In this retrospectively analysed study, the timing of the initial cranial ultrasound scan was not uniform. Variable scan timings would allow confounding by postnatal factors and would be expected to weaken the associations that we report. Alternatively, exposure to antenatal infection could increase the risk of postnatal adversity.

Females were over-represented among the infants with CDMI: a striking but unexplained finding. This finding could reflect a survival bias and indicate the importance of accounting for fetuses that are not admitted to the neonatal unit or who do not have an ultrasound scan. Some authors have advocated that time-oriented approaches to perinatal risk factors are adjusted for subsequent events using clearly defined markers [9]. This study illustrates a marker of brain injury that could be useful in time-oriented studies. However, our study highlights a significant issue with large-scale time-oriented studies performed in a routine clinical setting. Well-defined markers may not be applicable to all infants (e.g. the cases with “dubious” evidence of maternal infection).

These results provide proof of concept that the incidence of isolated PED on the initial cranial ultrasound is related to clinical evidence of maternal infection. The extremely wide confidence intervals suggest that this is a preliminary finding that may not be reliable. We have not found the association between in utero infection and IVH or parenchymal echolucencies that would have been expected from the literature, e.g. [10], [11], [12]. This discrepancy may be accounted for by factors such as differences in the ways infection is identified or because our study was too small to detect other associations. However, one explanation is that PED provide a more sensitive marker of in utero infection than other findings on cranial ultrasound. We speculate that the presence of PED identifies infants with early evidence of brain injury following exposure to in utero infection. Transient PED increase the risk of adverse long-term outcome but have relatively poor predictive value [2]. Thus, one group of infants with PED will not have permanent brain injury, while another group of infants with PED will have permanent brain injury. Studying the characteristics of these groups could indicate why some infants exposed to in utero infection develop neurodisability in later life but others do not [13]. In prospective studies, it would be interesting to account for the site and intensity of parenchymal echodensities.

Ultrasound is a less sensitive means to identify isolated parenchymal lesions than MRI [14], [15]. Small punctate lesions, especially away from the ventricles, are easily missed. This suggests that cranial ultrasound has a higher threshold for detecting brain injury than MRI. Despite this limitation, we found in the “real world” that PED are associated with CDMI. In contrast to MRI, cranial ultrasonography is widely available. Including PED (such as “flares”) in large-scale studies of how intra-uterine infection relates to brain injury is likely to contribute to an improved understanding of the pathways leading to childhood disability.

Acknowledgements

We are grateful to our consultant colleagues for their permission to access the clinical records of their patients and to all our colleagues for their care in recording the clinical events analysed in this study.

References

[1]. [1]Duggan PJ, Maalouf EF, Watts TL, Sullivan MHF, Counsell SJ, Allsop J, et al. Intrauterine T-cell activation and increased proinflammatory cytokine concentrations in preterm infants with cerebral lesions. Lancet. 2001;358:1699–1700. Abstract | Full Text | Full-Text PDF (117 KB) | CrossRef

[2]. [2]Dammann O, Leviton A. Duration of transient hyperechoic images of white matter in very-low-birthweight infants: a proposed classification. Dev Med Child Neurol. 1997;39:2–5. MEDLINE | CrossRef

[3]. [3]Paneth N. Classifying brain damage in preterm infants. J Pediatr. 1999;134:527–529. Full Text | Full-Text PDF (105 KB) | CrossRef

[4]. [4]Kuban K, Sanocka U, Leviton A, Allred EN, Pagano M, Dammann O, et al. White matter disorders of prematurity: association with intraventricular hemorrhage and ventriculomegaly. J Pediatr. 1999;134:539–546. Abstract | Full Text | Full-Text PDF (201 KB) | CrossRef

[5]. [5]The Developmental Epidemiology Network Investigators . The correlation between placental pathology and intraventricular hemorrhage in the preterm infant. Pediatr Res. 1998;43:15–19. MEDLINE

[6]. [6]Agarwal R, Chiswick M, Rimmer S, Taylor G, McNally R, Alston R, et al. Antenatal steroids are associated with a reduction in the incidence of cerebral white matter lesions in very low birthweight infants. Arch Dis Child Fetal Neonatal Ed. 2002;86:F96–F101. MEDLINE

[7]. [7]De Felice C, Toti P, Laurini RN, Stumpo M, Picciolini E, Todros T, et al. Early neonatal brain injury in histologic chorioamnionitis. J Pediatr. 2001;138:101–104. Abstract | Full Text | Full-Text PDF (66 KB) | CrossRef

[8]. [8]Spinillo A, Chiara A, Bergante C, Biancheri D, Fabiana D, Fazzi E. Obstetric risk factors and persistent increases in brain parenchymal echogenicity in preterm infants. BJOG: Intern J Obs Gyn. 2004;111:913–918.

[9]. [9]Dammann O, Leviton A, Bartels D, Dammann C. Lung and brain damage in preterm newborns: are they related? How Why?. Biol Neonate. 2004;85:305–313. MEDLINE | CrossRef

[10]. [10]Perlman J, Risser R, Boyles R. Bilateral periventricular leukomalacia in the premature infant: associated risk factors. Pediatrics. 1996;97:822–827.

[11]. [11]Vergani P, Patane L, Doria P, Borroni C, Cappellini A, Pezzullo J, et al. Risk factors for neonatal intraventricular haemorrhage in spontaneous prematurity at 32 weeks gestation or less. Placenta. 2000;21:402–407.

[12]. [12]Verma U, Tejani N, Klein S, Reale MR, Beneck D, Figueroa R, et al. Obstetric antecedents of intraventricular haemorrhage and periventricular leukomalacia in the low-birth-weight neonate. Am J Obstet Gynecol. 1997;176:275–281. Abstract | Full Text | Full-Text PDF (686 KB) | CrossRef

[13]. [13]Willoughby RE, Nelson KB. Chorioamnionitis and brain injury. Clin Perinatol. 2002;29:603–621. Full Text | Full-Text PDF (129 KB) | CrossRef

[14]. [14]Debillon T, N'Guyen S, Muet A, Quere M, Moussaly F, Roze J. Limitations of ultrasonography for diagnosing white matter damage in preterm infants. Arch Dis Child Fetal Neonatal Ed. 2003;88:F275–F279. MEDLINE

[15]. [15]Maalouf E, Duggan P, Counsell S, et al. Comparison of findings on cranial ultrasound and magnetic resonance imaging in preterm infants. Pediatrics. 2001;107:719–727.

a Academic Unit of Child Health, University of Manchester, UK

b Department of Obstetrics and Gynaecology, St. Mary's Hospital, Hathersage Road, Manchester M13 0JH, UK

c Department of Paediatrics, St. Mary's Hospital, Hathersage Road, Manchester M13 0JH, UK

d Department of Clinical Radiology, St. Mary's Hospital, Hathersage Road, Manchester M13 0JH, UK

e Department of Neonatal Medicine, St. Mary's Hospital, Hathersage Road, Manchester M13 0JH, UK

Corresponding author. University of Liverpool, Liverpool Women's Hospital, Crown Street, Liverpool L8 7SS, UK. Tel.: +44 151 702 4118; fax: +44 151 702 4024.

PII: S0378-3782(06)00152-6

doi:10.1016/j.earlhumdev.2006.05.007

12 of 12