Anatomy and pathophysiology of posthaemorrhagic ventricular dilatation

Cerebrospinal fluid (CSF) is mainly produced by the choroid plexus in the lateral ventricles and the roof of the third ventricle. Following a large IVH, multiple small blood clots throughout the ventricular system may clog up the channels of reabsorption of CSF. For some days after a large IVH, there may be a period of reduced CSF production but when CSF production returns, the lateral ventricles enlarge.

Over a period of weeks, chronic arachnoiditis may occur by which extracellular matrix proteins are laid down around the ependyma, the foramina of the fourth ventricle and in the subarachnoid space.1 The extracellular matrix proteins include fibronectin and laminin and are like cement or glue which can convert a potentially reversible CSF obstruction into a permanent CSF obstruction. There is evidence that transforming growth factor β (TGFβ) is involved in this process. TGFβ is the ‘fibrosis cytokine’ and upregulates the genes expressing extracellular matrix proteins. TGFβ is increased in the CSF of infants with posthaemorrhagic ventricular dilatation (PHVD) and those who go on to require ventriculoperitoneal (VP) shunt surgery have higher concentrations than those who do not.2 Similar findings have been observed in the CSF of adults with subarachnoid haemorrhage.3 A rat pup model of PHVD has demonstrated increased TGFβ, laminin and fibronectin in the ependyma and subependymal tissue.4 Transgenic mice that over-express TGFβ1 in the brain are born with hydrocephalus.5 However, other cytokines may be involved; in a study on hydrocephalic neonates, vascular endothelial growth factor was elevated in PHVD, whereas TGFβ was not.6

When haemoglobin is broken down, free iron is eventually liberated and analysis of CSF from infants with PHVD demonstrates free (non-protein-bound) iron in substantial amounts.7 Free iron is not found in normal CSF. Free iron is a potential source of free radicals and the presence of free iron in the centre of the immature brain for months may be an important mechanism of progressive white matter injury. Another important mechanism of white matter injury in premature infants is inflammation and CSF from infants with PHVD demonstrates high concentrations of several pro-inflammatory cytokines.8

The progressive accumulation of CSF changes the shape of the lateral ventricles from a slit to a balloon. The expanding ventricles distort the developing brain and pressure eventually starts to rise. Normal CSF pressure does not exceed 6 mm Hg. As the preterm skull is very compliant, the ventricles can expand without pressure rising initially but eventually pressure can rise to 10–15 mm Hg.9

Thus PHVD may produce progressive periventricular white matter injury over several months as a consequence of pressure, distortion, free radical injury and inflammation.

Diagnosis of PHVD including quantitative measurements

Levene has produced reference ranges for ventricular index according to gestational age.10 Since 1983, many centres in the UK have used 4 mm over the 97th centile for ventricular index as the ‘action line’ at which intervention should be considered (figure 1). This applies to both the right and the left ventricle. The ventricular index is measured from the falx to the lateral wall of the body of the ventricle as shown in figure 2. Sometimes, ventricles do not expand laterally, but become rounded or expand occipitally. Davies has produced reference ranges for anterior horn width (which is measured diagonally) and third ventricular width, measured in the coronal plane, as well as the thalamo-occipital dimension, measured in the sagittal plane, as shown in figures 2 and 3.11 To recognise this we have arbitrarily used the combination of three measurements—anterior horn width >4 mm (>1 mm over the 97th centile), thalamo-occipital dimension >26 mm (>1 mm over the 97th centile) and third ventricular width >3 mm (>1 mm over the 97th centile)—as an alternative definition for PHVD. Measurements on both sides must exceed these limits.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Reference values for ventricular width. The black line is 97th centile and the dotted line is 4 mm over the 97th centile (redrawn with permission from ref 10).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Cranial ultrasound scan showing moderate dilatation of the lateral ventricles. Coronal view with callipers showing anterior horn width (denoted by square calipers), third ventricular width (triangular calipers), and ventricular index (‘X’ and ‘+’ calipers) measurements.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Cranial ultrasound scan, sagittal view, showing considerable dilatation of the right lateral ventricles Callipers (‘+’) are showing the thalamo-occipital dimension.

Definition of excessive head enlargement

Head circumference enlarges by approximately 1 mm per day between 26 weeks of gestation and 32 weeks, and about 0.7 mm per day between 32 and 40 weeks.12 We regard a persistent increase of 2 mm per day as excessive. Measuring head circumference accurately, maximum fronto-occipital circumference, although low-tech, is not as easy as it sounds. Detecting a difference of 1 mm from day to day is difficult and we do not react to a difference of 2 mm from one day to the next unless there is other evidence of raised intracranial pressure. However, an increase of 4 mm over 2 days is more likely to be real and an increase of 14 mm over 7 days is definitely excessive.

Recognition of raised intracranial pressure

A change in the fontanelle from concave to bulging may be palpated but this is different from soft to tense which may be more ominous. The preterm skull is very compliant and can easily accommodate an increase in CSF by expanding with separation of the sutures. When CSF pressure was measured with an electronic transducer in infants expanding their ventricles after IVH, the mean CSF pressure was approximately 9 mm Hg, three times the mean in normal infants.9 There was a considerable range with some infants expanding their ventricles and heads at a pressure of 5–6 mm Hg and a small number with CSF pressure around 15 mm Hg. A CSF pressure of 9 mm Hg does not necessarily produce clinical signs but may be associated with an increase in apnoea or vomiting, hypotonia, hypertonia or decreased alertness. A structured neurological examination of the newborn such as that published by Dubowitz is recommended.13

Serial Doppler Resistance Index (RI) on the anterior cerebral artery is a useful and practical way of detecting impairment of cerebral perfusion by raised intracranial pressure and can easily be done during ultrasound imaging. RI is systolic velocity-diastolic velocity/systolic velocity. This measurement is independent of the angle of insonation. If intracranial pressure rises and arterial pressure does not rise, end-diastolic velocity decreases. In the context of PHVD and absence of a patent ductus arteriosus, RI increasing above 0.85 is suggestive of increasing pressure and RI of 1.0 indicates impaired perfusion. The sensitivity of the RI can be increased by applying pressure to the fontanelle during the examination. An infant who is close to the limit of cranial compliance responds by a large decrease in end-diastolic velocities, that is, an increase in RI.14 Serial amplitude-integrated electroencephalography may show deterioration (eg, increased discontinuity indicated by a decrease in the lower margin) as PHVD progresses and such changes may reverse with effective drainage.15

Prognosis of PHVD

The prognosis at diagnosis of PHVD using the above criteria is influenced by the presence of identifiable parenchymal lesions. The Ventriculomegaly and PHVD drug trials defined PHVD as a ventricular index at 4 mm above the 97th centile, as defined by Levene, and had standardised follow-up. If ultrasound examination shows no persistent echodensities or echolucencies (cysts), then approximately 40% of the children will develop cerebral palsy and about 25% will have multiple impairments.11 16 17

The presence of extensive haemorrhagic parenchymal infarction, in addition to PHVD, increases risk of cerebral palsy from around 40% to 80–90%, and large amounts of blood clot within the ventricles after IVH increase the risk of later shunt dependence.16 17 A Dutch retrospective study involving 144 surviving infants demonstrated lower rates of cerebral palsy, at around 10% for grade 3 and 50% for grade 4.18 This study also compared early CSF drainage, at a ventricular index of 4 mm above the 97th centile, to drainage beyond this threshold, and showed that early treatment was associated with better development quotient at 2 years and a decreased likelihood of shunt dependence; there was, however, no reduction in the rate of cerebral palsy with early drainage. A larger recent study involving 998 infants with birth weights <1000 g showed that, among those without haemorrhagic periventricular infarction (grade 4 IVH) 32% had a Bayley Mental Development Index (MDI) <50 and 39% had Psychomotor Developmental Index (PDI) <50. Among those with haemorrhagic periventricular infarction, 48% had MDI <50 and 65% had PDI <50.19 The rate of cerebral palsy was highest for grade 4 children with a shunt, at 80%.

Therapeutic interventions for PHVD

The following interventions have been used to treat PHVD but with no randomised trial evidence of efficacy and safety:

• ▶ repeated lumbar punctures (LPs) or ventricular taps

• ▶ repeated tapping through a ventricular access device

• ▶ drug treatment to reduce CSF production

• ▶ intraventricular fibrinolytic therapy

• ▶ external ventricular drain

• ▶ ventriculo-subgaleal shunt

• ▶ third ventriculostomy

• ▶ choroid plexus coagulation.

Repeated LP were suggested as a way of controlling pressure, preventing progressive ventricular enlargement and removing some of the red cells and protein from the CSF. Kressuer et al showed that a minimum of 10 ml/kg needed to be removed in order to have a significant effect on ventricular size.20 In contrast to older children and adults, ‘coning’ after LP in the presence of raised intracranial pressure is extremely rare in neonates. This is probably due to the pressure being only mildly raised in infants with open fontanelles and sutures. Having done hundreds of LPs in infants with slightly raised pressure, I have not seen any infant with PHVD cone. It is however, wise to limit the volume of CSF removed to a maximum of 20 ml/kg as larger volumes removed faster than 1 ml/kg/min may be followed by apnoea, bradycardia and desaturation. In our experience, only a minority of infants with PHVD have consistently communicating PHVD with a sufficient yield of lumbar CSF. Tapping of ventricular CSF has therefore to be considered. A policy of repeated early tapping lumbar or ventricular CSF for PHVD has been tested in four controlled clinical trials.21 Overall, there was no evidence that this approach reduced VP shunt surgery or disability and there was 7% infection among the infants who had repeated tapping in the ventriculomegaly trial.21

Diuretic drug treatment to reduce CSF production

Faced with this lack of effect and risk of infection from invasive procedures, pharmacological treatment to reduce CSF production seemed an excellent approach. Acetazolamide had been in clinical use for benign intracranial hypertension and appeared to have acceptable adverse effects as long as electrolyte and acid-base balance were monitored. Uncontrolled reports were positive about the effect of azetazolamide in PHVD. Further work showed that azetazolamide produced an initial increase in cerebral blood flow mediated by an increase in tissue CO₂ and inhibition of respiratory elimination of CO₂.22 Clinical investigation of infants with chronic lung disease of prematurity showed that acetazolamide produced an increase in pCO₂.23 Eventually a large multicentre randomised trial of acetazolamide combined with furosemide (which also reduces CSF production) was carried out. There was no clinical benefit and, to many neonatologists' surprise, the group receiving the combined drug treatment had significantly worse outcome in terms of shunt surgery and death or disability.17

Intraventricular fibrinolytic therapy

The idea of injecting a fibrinolytic agent intraventricularly grew out of Pang's PHVD model in which blood was injected intraventricularly into dogs; 80% developed hydrocephalus but if urokinase was injected intraventricularly, only 10% became hydrocephalic.24 Our own laboratory work showed that there was weak endogenous fibrinolytic activity in posthaemorrhagic CSF.25 A number of small non-randomised trials of intraventricular streptokinase, urokinase and tissue plasminogen activator (TPA), as well as two small randomised trials, have collectively shown that there is no reduction in VP shunt surgery and there is a risk of secondary intraventricular bleeding.26

Tapping CSF via a ventricular access device

The most widely used approach we have encountered in neonatal units who see a considerable number of infants with PHVD is the insertion of a ventricular access device, an Ommaya or Rickham reservoir (Codman and Shurtleff, Raynham, MA, USA), in those cases where repeated tapping is necessary to control excessive head enlargement and suspected raised pressure.27 This approximates the conservative or standard arm of the ventriculomegaly and PHVD drug trials and, in our, view, should be regarded as standard treatment now. Not all infants with PHVD demonstrate excessive head enlargement or signs of raised pressure so this is a selective approach. Once it becomes obvious that repeated CSF tapping is necessary, and a reservoir is inserted surgically, tapping can be carried out in a neonatal unit without neurosurgical staff. This avoids repeated taps through the fontanelle and intraparenchymal needle track injury. We withdraw 10 ml/kg of CSF over 10 min controlling the rate with a syringe. In centres with sufficient volume of patients, a ventricular access device can be inserted into extremely small infants (<1000 g) with a very low complication rate. In Bristol, a total of 93 such procedures have been carried out in preterm infants in the last 10 years with no cases of per-operative mortality, one infection, one CSF leak and one secondary IVH.

The standard threshold for intervention (width 97th centile + 4 mm) may be too late.28 We are currently participating in the Early versus Late Ventricular Intervention Study (ELVIS) which randomizes two treatment thresholds, the low (early) threshold being ventricular width over the 97th centile (with anterior horn width of 6 mm) and the higher (late) threshold being 4 mm over the 97th centile (with anterior horn width of 10 mm). Death or shunt dependence and disability at 2 years are the main outcomes. The coordinators are Dr Bert Smit, Sophie Children's Hospital, Rotterdam and Professor Linda de Vries, Utrecht University Medical Centre, Utrecht, Holland. Centres in Holland, UK, US and Sweden are taking part in ELVIS.

When is shunt surgery indicated in PHVD?

VP shunting should not be carried out as the primary treatment when progressive PHVD is diagnosed. There are several reasons for this. One is that the large amount of blood and protein in the CSF greatly increase the chance of the shunt system not draining well and requiring early revision.29 Second, ulceration of the skin over the valve site is likely as the skin of babies <1000 g in the first weeks is very thin and fragile with a consequent infection risk. Finally, a VP shunt has its own long-term complications and approximately 50% of infants with PHVD do not need a permanent shunt. In our practice VP shunting is done at around term if there is persistent need for tapping to maintain normal head growth or if there is persistent excessive head enlargement. If an infant has had a reservoir inserted and is having repeated taps to maintain normal head growth, this is continued until CSF protein is <1.5 g/l and the infant is free from infection with an acceptable weight (usually 2.5 kg in Bristol but could be lower in individual cases). When these conditions are met, tapping stops and the head circumference is measured daily. If the head circumference increases by 2 mm/day, ultrasound is used to confirm that the increase is CSF and not brain growth. Insertion of a VP shunt is indicated if the excessive growth persists over several days in the above circumstances. Some infants grow at a rate just below 2 mm/day, but do so persistently and cross all the centiles, eventually developing an inappropriately large head for the body.

External ventricular drainage, third ventriculostomy and choroid plexus coagulation have been carried out in infants with PHVD but without any controlled trials.

Drainage, irrigation and fibrinolytic therapy

Drainage, irrigation and fibrinolytic therapy (DRIFT) is an approach that grew out of the unsatisfactory results of the above treatments. The objective is to reduce pressure and distortion early and to remove pro-inflammatory cytokines and free iron from within the ventricles.30 The procedure involves insertion of right frontal and left occipital ventricular catheters. TPA is injected intraventricularly at a dose that is insufficient to produce a systemic effect and this is left for approximately 8 h. Artificial CSF (Torbay Pharmaceutical, Paignton, UK) is then pumped into the frontal ventricular catheter at 20 ml/h with continuous intracranial pressure monitoring. The occipital ventricular catheter is simultaneously connected to a sterile closed ventricular drainage system and the height of the drainage reservoir adjusted to increase or decrease drainage to maintain intracranial pressure below 7 mm Hg and net loss of 60–100 ml CSF/day in 48 h. The drainage fluid initially looks like cola but gradually clears to look like white wine, at which point irrigation is stopped and the catheters removed. This commonly takes 72 h but can be up to 7 days. Figure 4 shows MR imaging documenting the removal of intraventricular debris with this procedure.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

T2 weighted axial MR images before DRIFT (on the left) and after DRIFT (on the right). Before DRIFT there was extensive debris within the dilated ventricles and oedema of the left hemisphere. After DRIFT, nearly all the debris had been removed and the hemispheric oedema had reduced. (with permission from Whitelaw et al30).

DRIFT has been tested in a randomised clinical trial that recruited 77 infants in four centres. Although DRIFT did not significantly lower the need for shunt surgery, severe cognitive disability at 2 years Bayley (MDI <55) was significantly reduced.31 Median MDI was improved by more than 18 developmental points in the DRIFT group.32 This is the first intervention demonstrated in a randomised trial to improve outcome in infants with PHVD. However, the DRIFT trial was stopped before full planned recruitment and the study group is relatively small. Furthermore, the DRIFT procedure, as carried out in the trial, is very difficult because of the need to monitor intracranial pressure reliably throughout and the need to respond to variations in the amount of fluid draining and to flush out occasional complete blockages. The need to maintain a perfect sterile technique to avoid infection is another challenge. We now refer to this technique as ventricular lavage and are working on minor modifications such as wider bore catheters to improve ease of use, efficacy and safety before its wider use can be encouraged. TPA is a two-edged sword. On the one hand, there is evidence of increased secondary bleeding but on the other hand, observational evidence that, if there is no secondary bleed, TPA helps to mobilise blood clot to be removed. One can, quite legitimately, weigh these two arguments differently. In the Pediatrics paper, we came down on the side of trying to reduce secondary IVH by restricting TPA.31 However, there may be benefit for more infants if TPA is given routinely, accepting that a minority will have secondary bleeding which can be managed.

Thus our understanding of disease mechanisms in PHVD has advanced and testing of interventions continues. Centralisation of infants with PHVD to large neonatal centres with integrated neurosurgery is important if further advances are to be made.