Neurodevelopmental outcome at 2 years after

CLINICAL ARTICLEJ Neurosurg Pediatr 26:495–503, 2020

Despite continuous progress in neonatal care, a rel-evant rate of premature infants (25%–30%) still develop intraventricular hemorrhage (IVH).1,2 In

most cases, IVH in neonates is caused by germinal matrix rupture within the first 72 hours after preterm delivery, which leads to a primary destruction of neuronal and glial

precursor cells in the parenchyma to a varying extent.3 Ac-cording to the Papile classification, IVH is classified into the following categories: grade I, with isolated subepen-dymal hemorrhage without ventricular involvement; grade II, with IVH including ≤ 50% of ventricular volume; grade III, with IVH extending into > 50% of the ventricular vol-

ABBREVIATIONS AED = antiepileptic drug; BSID II MDI = Bayley Scales of Infant Development, 2nd Edition, Mental Developmental Index; CI = confidence interval; CrI = credible interval; CSF = cerebrospinal fluid; DRIFT = drainage, irrigation, and fibrinolytic therapy; GMFCS = Gross Motor Function Classification System; IQR = interquartile range; IVH = intraventricular hemorrhage; NEL = neuroendoscopic lavage; OR = odds ratio.SUBMITTED March 23, 2020. ACCEPTED May 11, 2020.INCLUDE WHEN CITING Published online August 7, 2020; DOI: 10.3171/2020.5.PEDS20211.* A.A. and U.W.T. contributed equally to this study.

Neurodevelopmental outcome at 2 years after neuroendoscopic lavage in neonates with posthemorrhagic hydrocephalus*Philine Behrens, cand med,1 Anna Tietze, Dr med,2 Elisabeth Walch, Dr med,3 Petra Bittigau, PD, Dr med,3 Christoph Bührer, Dr med,4 Matthias Schulz, PD, Dr med,1 Annette Aigner, PhD,5 and Ulrich-Wilhelm Thomale, Prof Dr med1

1Pediatric Neurosurgery, 2Neuroradiology, 3Pediatric Neurology, 4Neonatology, and 5Institute of Biometry and Clinical Epidemiology, Charité—Universitätsmedizin Berlin, Germany

OBJECTIVE A standardized guideline for treatment of posthemorrhagic hydrocephalus in premature infants is still miss-ing. Because an early ventriculoperitoneal shunt surgery is avoided due to low body weight and fragility of the patients, the neurosurgical treatment focuses on temporary solutions for CSF diversion as a minimally invasive approach. Neuro-endoscopic lavage (NEL) was additionally introduced for early elimination of intraventricular blood components to reduce possible subsequent complications such as shunt dependency, infection, and multiloculated hydrocephalus. The authors report their first experience regarding neurodevelopmental outcome after NEL in this patient cohort.METHODS In a single-center retrospective cohort study with 45 patients undergoing NEL, the authors measured neu-rocognitive development at 2 years with the Bayley Scales of Infant Development, 2nd Edition, Mental Developmental Index (BSID II MDI) and graded the ability to walk with the Gross Motor Function Classification System (GMFCS). They further recorded medication with antiepileptic drugs (AEDs) and quantified ventricular and brain volumes by using 3D MRI data sets.RESULTS Forty-four patients were alive at 2 years of age. Eight of 27 patients (30%) assessed revealed a fairly normal neurocognitive development (BSID II MDI ≥ 70), 28 of 36 patients (78%) were able to walk independently or with minimal aid (GMFCS 0–2), and 73% did not require AED treatment. Based on MR volume measurements, greater brain volume was positively correlated with BSID II MDI (rs = 0.52, 95% CI 0.08–0.79) and negatively with GMFCS (rs = −0.69, 95% CI −0.85 to −0.42). Based on Bayesian logistic regression, AED treatment, the presence of comorbidities, and also cerebel-lar pathology could be identified as relevant risk factors for both neurodevelopmental outcomes, increasing the odds more than 2-fold—but with limited precision in estimation.CONCLUSIONS Neuromotor outcome assessment after NEL is comparable to previously published drainage, irrigation, and fibrinolytic therapy (DRIFT) study results. A majority of NEL-treated patients showed independent mobility. Further validation of outcome measurements is warranted in an extended setup, as intended by the prospective international multicenter registry for treatment of posthemorrhagic hydrocephalus (TROPHY).https://thejns.org/doi/abs/10.3171/2020.5.PEDS20211KEYWORDS posthemorrhagic hydrocephalus; intraventricular hemorrhage; neuroendoscopic lavage; neonate

J Neurosurg Pediatr Volume 26 • November 2020 495©AANS 2020, except where prohibited by US copyright law

Unauthenticated | Downloaded 06/08/22 04:05 PM UTC

Behrens et al.

J Neurosurg Pediatr Volume 26 • November 2020496

ume; and grade IV, in which parenchymal hemorrhage is additionally observed.4,5

In general, the outcome of patients is mainly deter-mined by possible comorbidities of prematurity, such as the extent of bleeding resulting in possible secondary brain damage that includes the development of posthemorrhagic ventricular dilatation or even posthemorrhagic hydroceph-alus.6,7 Hydrocephalus causes white matter shearing forces by massive enlargement of the ventricles, and consecutive intracranial pressure may result in compression-related brain tissue injury.6 Shunt dependency is observed in 38%–87% of cases and may be associated with secondary treatment complications.1 In addition, the accumulation of intraventricular blood triggers an immune response re-cruiting cytotoxic substances, which might cause further neuronal damage.8

The primary focus of neurosurgical treatment of IVH-related hydrocephalus in this patient cohort is the diver-sion of CSF by a temporizing measure, because an early ventriculoperitoneal shunt implantation may be associated with higher shunt failure rates due to low body weight and the amount of blood and its degradation products in the CSF.9,10 Different approaches, directed toward reduction of the shunt dependency rate and improved neurological out-come (e.g., by repeated lumbar or ventricular tapping or the use of acetazolamide and furosemide to reduce CSF pro-duction), did not yield any beneficial effect.10 In a clinical trial comparing ventricular access devices and ventriculo-subgaleal shunts, no differences were found with respect to shunt infection, surgical revision, or mortality rate.9 These results were reinforced by a meta-analysis showing a simi-lar range of outcomes for ventriculosubgaleal shunts and ventricular access devices regarding obstruction, infection, revision, arrest of hydrocephalus, good neurodevelopmen-tal outcome, and death. Interestingly, external ventricular drain implantation showed a lower infection and higher ar-rested hydrocephalus rate compared with the other meth-ods, but also a relevant higher surgical revision rate.1

More invasive methods were directed at the early elimination of blood products from the ventricles. The ef-ficacy of the drainage, irrigation, and fibrinolytic therapy (DRIFT) study to prevent hydrocephalus through early removal of intraventricular blood, cytokines, and iron re-sulted in no differences between groups regarding shunt dependency or death, but a 35% versus 8% rate of second-ary hemorrhage in the standard versus the DRIFT group.10 In addition, an analysis performed 2 years after inclusion showed a reduction of severe cognitive disability indicated by the Bayley Scales of Infant Development, 2nd Edition, Mental Developmental Index (BSID II MDI) and a non-significantly lower rate of epilepsy in the DRIFT group.7 The 10-year follow-up further confirmed improved cogni-tive function in the DRIFT group compared with the stan-dard group.11

Neuroendoscopic lavage (NEL), on the other hand, rep-resents a single intervention procedure, is minimally inva-sive, and is a far more controlled technique to achieve the elimination of intraventricular blood products, because it obviates the use of fibrinolytic agents, thereby decreas-ing the risk for secondary hemorrhage.12 Its feasibility has been confirmed in previous studies and a reduction of

shunt frequency, CSF infections, and the development of a multiloculated hydrocephalus were shown.2,4,12 However, developmental outcome measures are still missing for the assessment of NEL. Thus, the aim of this research was to evaluate the neurodevelopmental and motor outcomes, as well as the ventricular and brain volume measurements 2 years after NEL in a single-center retrospective study. Potential clinical risk factors for impaired neuromotor outcomes were investigated.

MethodsThis retrospectively analyzed cohort study was ap-

proved by the institutional research ethics board. In a data-base review we included all consecutively treated patients who underwent the intervention of NEL at our institution between September 2010 and May 2016 (n = 45). The in-dication for surgery and the surgical technique were de-scribed previously4,12 and are described more in detail in the supplemental material.

Clinical baseline data were derived from the in-house database, such as gestational age, body weight at birth, IVH grade, comorbidities, time from hemorrhage to NEL, and the use of antiepileptic drugs (AEDs). We included only patients who were alive at the intended follow-up of 2 years, and thereby excluded 1 patient who died from bronchopulmonary dysplasia at the age of 2 months and 3 weeks. At the follow-up of 2 years corrected age, taking the calculated birthday after premature birth into account, patients’ families visited the hospital again and we per-formed neurodevelopmental assessment by using BSID II MDI, the Gross Motor Function Classification System (GMFCS), and MRI as described below.

BSID II MDIThe BSID II MDI score was assessed in the Pediatric

Care Center for Complex Chronic Diseases. The BSID II MDI is designed to measure the developmental progress of infants between 1 month and 42 months. It helps to quantify the following aspects in a playful and interactive examination: memory, learning ability, problem-solving, early concept of numbers, generalization, categorization, language, and communicative competences. According to the developmental classification, scores ≥ 85 indicate a normal range of development, whereas scores between 70 and 84 indicate the risk for subsequent developmental delay, and values < 70 indicate a manifest delay in devel-opment. Scores from 50 to 55 indicate severe cognitive disabilities, whereas values < 50 are not measurable.

GMFCSPossible cerebral palsy is quantified by evaluating the

ambulatory ability in the Pediatric Care Center for Com-plex Chronic Diseases by using the GMFCS, which fo-cuses on self-initiated movement.13,14 According to the expanded and revised version of the GMFCS, 5 different levels of cerebral palsy (1–5) are differentiated and are ad-ditionally adapted to the specific age. Because the infants of this study population were assessed at 2 years, the in-dicative factors for a child’s movement between the 2nd and 4th birthday were evaluated (Table 1).


https://thejns.org/doi/suppl/10.3171/2020.5.PEDS20211

J Neurosurg Pediatr Volume 26 • November 2020 497

Behrens et al.

Ventricular and Total Intracranial Brain VolumeThe ventricular and brain volumes were quantified by

using a semiautomated CSF and brain tissue segmentation tool (ITK-SNAP; an open software program that can be downloaded at http://www.itksnap.org15). High-resolution 3D T1-weighted MRI data as part of the clinical MRI pro-tocol were acquired on a 1.5-T system (Genesis Signa; GE Healthcare) until 2015 and subsequently on a 3-T system (Magnetom Skyra; Siemens Healthineers).

For tissue-type segmentation (i.e., the quantification of brain parenchyma and the ventricular system), 3D T1-weighted MRI data were loaded into the ITK-SNAP tool and tissue classes were manually defined using a built-in presegmentation step. At least 10 seed points were then placed into the region of interest (CSF and brain tissue) and the evolution tool was used to let these seeds automatically evolve or contract to include CSF and brain tissue, respec-tively. Either incomplete segmentation or oversegmenta-tion occurred in all cases, and the volumes were manually corrected slice by slice. Finally, the volume calculation tool was used to compute the number of voxels and, consider-ing the voxel size, the volume was calculated (Fig. 1).

Statistical AnalysisWe report absolute and relative frequencies for categor-

ical variables and median with interquartile range (IQR) for ordinal and continuous variables, for the whole cohort, as well as neurodevelopmental and motor outcomes. As-sociations among brain volume, ventricular volume, and neurodevelopment were assessed with Spearman correla-tion coefficients, reported along with 95% confidence in-tervals (CIs) calculated via z-transformation. Risk factors for worse neurodevelopment, defined as BSID II MDI < 70 or GMFCS > 2, were investigated with Bayesian lo-gistic regression models with a weakly informative prior distribution to deal with complete separation,16 i.e., the sit-uation in which combinations of variables are not present or occur very rarely in the data. Because missing follow-up data are expected to be associated with the condition of a patient, these analyses were performed on multiply

imputed data for all patients present at follow-up (i.e., 42 observations). Subsequently, the odds ratio (OR) estimates and 95% credible intervals (CrIs) were pooled by drawing 100,000 samples from each posterior distribution of the coefficients of all imputed data sets and deriving the mean and 2.5% and 97.5% quantiles. Based on the same analy-ses, we also report unadjusted effect estimates as sensitiv-ity analyses and assess risk factors for shunt dependency. Due to the small sample size and the issue of complete separation, all results have to be interpreted with care, and the focus should be on effect estimates rather than on sta-tistical significance, where variables changing the odds of the outcome by more than 50% are deemed relevant in this setting.

Statistical analyses were performed using R software,17 as well as additional R packages for data handling and plotting,18 Spearman correlation,19 Bayesian logistic re-gression,20 and multiple imputation,21 where the latter was based on the classification and regression tree algorithm with 100 imputed data sets.

ResultsPatient Characteristics

Of 44 eligible patients, 2 were lost to follow-up. Of the 42 patients with at least one 2-year follow-up appointment, 36 were assessed with GMFCS, 27 with BSID II MDI, and 29 received ventricular and supratentorial volume measurements on MRI volume data sets acquired at a me-dian of 25.1 months (IQR 22.3–30.3 months) after surgery. The information about AED therapy was available based on hospital records for 41 patients. Thirty of 41 patients (73%) did not require AED treatment. Basic patient char-acteristics are given in Table 2.

Basic clinical data from missing patients showed that the 2 patients who were lost to follow-up had a higher birth weight than those with some follow-up appointments (1465 g vs 1168.5 g), and both were classified with an IVH

TABLE 1. Evaluation of GMFCS between 2nd and 4th birthday in patients with neonatal posthemorrhagic hydrocephalus

Level Abilities

0 • Free walking without spasticity1 • Free walking, moving in and out of sitting position without

assistance2 • Walking with limitations/preferred with assistance holding

on to furniture• Moving in and out of sitting without assistance

3 • Walking with handheld mobility device• W-sitting, assuming of sitting position with adult support

4 • No walking, powered mobility required• Trunk support required for sitting• Creeping or crawling

5 • No independent movement• Limited head and trunk control

Based on data from Palisano and colleagues.13,14

FIG. 1. Illustration of CSF and brain tissue segmentation. Figure is avail-able in color online only.


Behrens et al.


grade of IV. In our final cohort of 42 patients, 6 patients received systemic antibiotics due to increased infectious parameters in serum. Of those, 2 patients showed positive germ cultures in CSF. For 3 of them, the BSID II MDI score was missing but not the GMFCS score. Patients without a BSID II MDI assessment tended to have lower GMFCS values (median GMFCS of level 0 vs level 2) and a lower body weight at NEL (1375 g vs 1615 g), and their proportion of IVH grade IV was lower (40% vs 56%), as was the proportion of shunt dependency (47% vs 67%). Patients without a GMFCS assessment had similar birth

weight (1174 g vs 1169 g), but lower body weight at NEL (1338 g vs 1630 g). Their proportion of IVH grade IV was higher (66.7% vs 47.2%), as was the proportion of comor-bidities (50.0% vs 27.8%). Forty percent of those without a BSID II MDI assessment also did not have a GMFCS assessment (Supplemental Table S1).

Outcome MeasuresOf the 36 patients assessed with the GMFCS, 13 (36%)

showed no cerebral palsy—representing a normal gait. They were classified with level 0 in the Spearman cor-

TABLE 2. Patient characteristics by outcome parameter at 2-year follow-up

Characteristic Total, n = 42BSID II MDI Score GMFCS Score

≥70, n = 12 <70, n = 15 ≤2, n = 28 >2, n = 8

Gestational wk Median (IQR) 27 (25–32) 28 (26.8–39) 26 (24–30) 30 (26–35.2) 24.5 (23.8–27)Birth weight in g Median (IQR) 1168.5 (809.5–2082.5) 1275 (844–3276.2) 930 (711.5–1350) 1208.5 (847.5–2822.5) 849.5 (601.8–1330)IVH grade II 4 (10%) 2 (17%) 0 (0%) 4 (14%) 0 (0%) III 17 (40%) 4 (33%) 6 (40%) 13 (46%) 2 (25%) IV 21 (50%) 6 (50%) 9 (60%) 11 (39%) 6 (75%)Wks from hemorrhage to NEL Median (IQR) 2.6 (2–3.7) 2.6 (2.2–3.7) 3 (2.4–3.8) 2.6 (2–3.4) 2.7 (2.4–4.4)Body weight in g at NEL Median (IQR) 1610 (1248.8–2406.2) 1690 (1473.8–3323.8) 1580 (1286–1922.5) 1630 (1256.2–3027.5) 1642.5 (1390–1913.8)Initial infection No 36 (86%) 10 (83%) 14 (93%) 23 (82%) 7 (88%) Yes 6 (14%) 2 (17%) 1 (7%) 5 (18%) 1 (12%)VP shunt No 17 (40%) 4 (33%) 5 (33%) 11 (39%) 3 (38%) Yes 25 (60%) 8 (67%) 10 (67%) 17 (61%) 5 (62%)Shunt revision No 12 (48%) 3 (38%) 6 (60%) 7 (41%) 3 (60%) Yes 13 (52%) 5 (62%) 4 (40%) 10 (59%) 2 (40%)Ventricular vol in ml Median (IQR) 170.9 (94.3–226.2) 123.9 (94.3–225.2) 218.4 (172.6–339) 123.9 (60.1–208.5) 301.4 (221.4–491)Brain vol in ml Median (IQR) 932.7 (739–1020.5) 939.8 (838.7–1041.1) 693.3 (645.1–910.8) 962.7 (906.9–1036.5) 653 (621.8–704.8)Cerebellar pathology None 36 (86%) 12 (100%) 10 (67%) 26 (93%) 4 (50%) Isolated 4th V 3 (7%) 0 (0%) 3 (20%) 1 (4%) 2 (25%) Cerebellar atrophy* 6 (14%) 0 (0%) 5 (33%) 2 (7%) 4 (50%)AEDs No 30 (71%) 11 (92%) 7 (47%) 23 (82%) 3 (38%) Yes 11 (26%) 1 (8%) 8 (53%) 5 (18%) 5 (62%)Comorbidities No 29 (69%) 12 (100%) 7 (47%) 24 (86%) 2 (25%) Yes 13 (31%) 0 (0%) 8 (53%) 4 (14%) 6 (75%)

VP = ventriculoperitoneal; 4th V = fourth ventricle.Unless otherwise indicated, values are expressed as the number of patients (%).* Two of 6 patients with cerebellar atrophy also had isolated fourth ventricle.




Behrens et al.

relation analysis (Table 3). Four patients (11%) were able to walk freely with spasticity (level 1); 11 patients (31%) preferred to walk with assistance (level 2); 3 patients (8%) used a handheld mobility device (level 3); and another 3 patients (8%) tended to creep and crawl or needed pow-ered mobility and trunk support while sitting (level 4). Only 2 patients (6%) were not able to move independently (level 5) (Fig. 2).

Of 27 patients assessed with BSID II MDI, 8 patients (30%) reached a score ≥ 85, which indicates a neurocog-nitive development within normal range at the corrected age of 2 years. Four patients (15%) showed minor develop-mental delay, as indicated by a score between 70 and 84, whereas 3 (11%) presented with moderate developmental delay at a score range from 55 to 69, and 12 children (44%) had severe cognitive disability, indicated by a score < 55 at the time of investigation (Fig. 2).

Brain and Ventricular VolumeThe median brain volume of the 29 patients was 932.7

ml (IQR 739–1020.5 ml) and the median ventricular vol-ume was 170.9 ml (IQR 94.3–226.2 ml). Brain volume measures showed a positive correlation with BSID II MDI and a negative one with GMFCS; the association appears stronger with GMFCS. The association of both BSID II MDI and GMFCS scores with ventricular volume is weak-er compared with their association with brain volume (Fig. 3, Table 3). On MRI no multiloculated hydrocepha-lus with CSF communication disturbances was observed. However, 3 patients developed an isolated fourth ventricle and were treated with endoscopic stented aqueductoplasty. In addition, we observed 6 patients with severe cerebellar atrophy, of whom 2 had an isolated fourth ventricle.

Neurodevelopmental and Motor OutcomesBased on 42 patients for whom some follow-up was

observed, older gestational age, time from hemorrhage to NEL, and also shunt dependency had negligible asso-ciations with the neurodevelopmental and motor outcome measures used after 2 years. An IVH grade of III or IV might have a relevant detrimental effect regarding both outcomes, but given the few observations in patients with grade II, this effect could not be estimated with sufficient precision in any model.

The use of AEDs, the presence of comorbidities, and also cerebellar pathology could be identified as relevant risk factors for both neurodevelopmental and motor out-comes. Adjusting for all other factors, AEDs increased the odds of worse neurocognitive and ambulatory outcomes approximately 3-fold (OR 3.06, 95% CrI 0.52–20.53 and OR 2.68, 95% CrI 0.46–16.69, respectively), and comor-

bidities increased the odds approximately 2.5-fold (OR 2.45, 95% CrI 0.34–24.24 and OR 2.47, 95% CrI 0.36–16.77, respectively). Without adjustment for confounding, these effects were even stronger and could be estimated with higher precision, resulting in smaller CrIs. A cerebel-lar pathology independently increases the odds of worse neurocognitive and ambulatory outcomes almost 2-fold (OR 1.92, 95% CrI 0.16–48.12 and OR 1.69, 95% CrI 0.25–11.59, respectively), although with very low preci-sion in estimation. Without adjustment, these effects are more pronounced with OR estimates above 5—but still with high uncertainty in the actual effect (Fig. 4).

Shunt DependencyIn the cohort, 26 of 44 patients (59%) became shunt

dependent. Time from hemorrhage to NEL was not rel-evantly lower for patients without a shunt (median age 2.4 weeks [IQR 1.9–3.1 weeks] vs 3.1 weeks [IQR 2.3–3.8

TABLE 3. Comparisons between neurodevelopment, ventricular volume, and brain volume in patients with hydrocephalus

Outcome Measure GMFCS Score Ventricular Vol Brain Vol

BSID II MDI −0.69 (−0.85 to −0.42) −0.37 (−0.7 to −0.1) 0.52 (0.08 to 0.79)GMFCS score 0.51 (0.15 to 0.76) −0.69 (−0.85 to −0.40)Ventricular vol −0.47 (−0.72 to −0.13)

Values are expressed as the Spearman correlation coefficient (95% CI).

FIG. 2. Stacked bar chart showing the relative distribution in as-sessed outcome for BSID II MDI neurodevelopmental score (n = 27) and GMFCS score (n = 36), respectively, during 2 years of follow-up. GMFCS 0 = no spasticity. Figure is available in color online only.


Behrens et al.


weeks]; Supplemental Fig. S1B). Adjusting for baseline pa-rameters (gestational week, IVH grade, comorbidities, and infection), the time from hemorrhage to NEL, and also the gestational age and comorbidities had negligible effects on the outcome. A higher IVH grade might increase the odds of shunt dependency relevantly (OR 4.45, 95% CrI 0.55–36.29), as might the presence of infections at baseline (OR 4.16, 95% CrI 0.58–29.70) (Supplemental Fig. S1A).

DiscussionAfter introducing NEL in our department in 2010 as

a routine technique for treating decompensated posthem-orrhagic hydrocephalus in neonates, we herewith present functional outcome measurements as relevant parameters to validate the effectiveness of such a technique for the first time. Almost one-third of children in our study who underwent NEL after IVH showed a neurocognitive devel-opment within normal range, and three-quarters showed the ability to walk freely or with minor aid. More than two-thirds did not suffer from treatment-dependent epi-lepsy. However, 44% of the infants had a severe cognitive disability and more than 1 in 5 patients could not walk or constantly needed a mobility device. We found a positive correlation between brain volume and ambulatory abil-ity, which was less pronounced for the cognitive outcome. Ventricular volume measurements correlated less clearly with outcome parameters compared with brain volume measurements. In logistic regression analysis, the treat-ment with AEDs, the presence of comorbidities at base-line, and a cerebellar pathology were associated with im-paired outcome after 2 years.

The indication for intervention in our cohort was sim-ilar to the inclusion criteria of the DRIFT study.7 A di-rect comparison of the baseline parameters at the time of treatment is presented in Table 4. There is no indication

of major differences, but there was a tendency for body weight in our cohort to be higher at NEL. Because a more detailed comparison of the two studies is not possible, we can only speculate about the reasons for the different out-comes in the NEL and DRIFT groups. One explanatory factor might be the single intervention in the NEL group in a controlled environment of the operating room, as op-posed to repeated procedures performed over several days by changing staff on the neonatal intensive care unit in the DRIFT group. This safety measure is mainly reflected in a higher rate of secondary IVH, which was 35% in the DRIFT group compared with 8% in the NEL group. It is, however, not confirmed by either the mortality rate (6% vs 5%) or the infection rate—the latter was 0% in the DRIFT group compared with 3.6% in a bicenter experience in which the NEL technique was used.4

Comparing the BSID II MDI from our NEL study co-hort with the results of the previously published random-ized controlled DRIFT study,7 fewer surviving children in the DRIFT group had a normal cognitive development (BSID II MDI score ≥ 85) 2 years after intervention than in our NEL group (23% vs 30%). Also, more children showed minor (BSID II MDI score 70–84) and moder-ate (BSID II MDI score 55–69) developmental delays in the DRIFT group (minor: 26% vs 15%; moderate: 20% vs 11%). However, only 31% of the surviving children in the DRIFT group suffered from severe cognitive disability (BSID II MDI score < 55) compared with 44% in the NEL group. Taking into account the higher mortality of 6% in the DRIFT group compared with 2% in our present NEL cohort, we may summarize that the DRIFT outcome re-sults are not relevantly different from the results presented here.10 The comparison of neurocognitive and ambulatory outcomes after NEL with those reported for the control group of the DRIFT study (BSID II MDI score ≥ 85: 28% vs 30%; BSID II MDI score 70–84: 9% vs 15%; BSID II

FIG. 3. Scatterplots of brain volume versus BSID II MDI neurodevelopmental score (left) and GMFCS score (right). GMFCS 0 = no spasticity. (For Spearman correlation coefficient, see Table 3.)





Behrens et al.

MDI score 55–69: 3% vs 11%; BSID II MDI score < 55: 59% vs 44%) indicates a benefit from NEL for avoiding severe disability.

Results regarding ambulatory function are more dif-ficult to compare with the DRIFT group, in which the GMFCS was not used. However, the ability to walk was described to be normal in 17% of the DRIFT group, which might correspond to the 36% of NEL patients without spasticity. Similarly, 44% in the DRIFT group were de-scribed as unable to walk without assistance, whereas the remaining patients had normal or abnormal gait with re-duced mobility.7 This subgroup might best be compared with GMFCS 3–4 in our study, in which it represents 22%

of the NEL cohort. Overall, this indicates that in this sin-gle-center study NEL resulted in better gait ability after 2 years; however, such a conclusion must be interpreted cau-tiously due to different study setups. Interestingly, in the DRIFT study there were no clear differences with respect to gait ability in the DRIFT cohort compared with the pa-tients treated with standard care (normal or nonfluent gait 42% vs 45%; reduced or no mobility 58% vs 55%).

The rate of shunt dependency in the DRIFT study was rather low compared with our cohort after applying NEL (38% vs 60%). Here it is important to state that the in-dication for shunt insertion remains a subjective decision made mainly by the treating neurosurgical team, because

FIG. 4. OR estimates with 95% CrIs based on Bayesian multiple (A) and univariate (B) logistic regression models with multiply imputed data. Outcome parameters are neurodevelopmental score BSID II MDI < 70 and GMFCS > 2. Figure is available in color online only.


Behrens et al.


all states of marginally compensated hydrocephalus may be left without shunting and will eventually reach a steady state. This makes shunt independence a fragile parame-ter; some patients with a shunt might end up with better outcomes given that brain parenchyma can be preserved by avoiding high intracranial pressure periods compared with patients without a shunt. Nevertheless, it is debatable whether low gestational age as well as the presence of an early infection is associated with higher odds of shunt de-pendency.4 These factors may be taken into account when deciding about shunt surgery and consulting the parents.

The DRIFT study did not measure brain or ventricu-lar volume as outcome parameters. There are only limited data in the literature about brain volume development at 2 years after neonatal posthemorrhagic hydrocephalus. The technique of automated volume segmentation for hydroce-phalic radiographic imaging was reported previously.15,22–25 In a mixed hydrocephalus cohort in which endoscopic third ventriculostomy and choroid plexus cauterization (ETV-CPC) were used for treatment, after 12 months a median CSF volume of 311 ml was measured, which is relevantly larger than the values in our cohort. Brain vol-ume measurements are only presented as relative values in this study.24 It was shown that decreased CSF volume is observed in patients with shunts; however, a correlation of brain volume with outcome was not investigated. The same study did not find significant differences in outcome between individuals with and without shunts.

Previous studies have shown that IVH and posthemor-rhagic hydrocephalus are highly correlated with deep gray matter and cerebellar volume, as well as with presumed microstructural white matter integrity,26 and that periven-tricular leukomalacia in infants is associated with their visual function.27 In fact, we found a stronger correlation between brain tissue volume and motor and developmen-tal outcome than between ventricular volumes and these outcome measures. Studies of normative brain volume growth have determined the brain volume at 2 years to be between 900 and 1350 ml.28 In spite of the prematurity and various comorbidities in our cohort, approximately 50% of patients have brain volumes above the 3rd percentile as de-fined by Peterson and colleagues. Our data may underline the importance of brain tissue volume measurement rather than ventricular size as an outcome surrogate, because the

latter will not detect any brain volume changes indirectly when external CSF spaces are additionally enlarged. Fur-ther investigations are necessary to verify the importance of brain volume measures in relation to neurodevelopmen-tal outcome.

Limitations of the StudyThe low prevalence of neonatal posthemorrhagic hy-

drocephalus is reflected by the small sample size in our study. Because we rely on data from a retrospective co-hort, in which outcome parameters were measured on a routine basis depending on whether families actually re-ceived follow-up at our institution, we were not able to col-lect all information on outcome parameters after 2 years in the NEL cohort. Due to the fact that, first, measurements of BSID II MDI scores, GMFCS scores, ventricular vol-ume, and brain volume are correlated with each other to a moderate to high extent, and, second, baseline information was associated with whether or not observations of the outcome parameters were missing at follow-up, a multiple imputation approach appears to be appropriate and uses the available information most efficiently. However, results have to be interpreted with care; e.g., we acknowledge that patients with missing GMFCS scores tended to have higher IVH grades and more comorbidities, potentially re-sulting in worse ambulatory outcome. Furthermore, brain volume measurements will not detect any microstructural changes that also may be linked to posthemorrhagic and hydrocephalus-derived brain injury.

Regarding the validity of the neurocognitive evalua-tion by BSID II MDI, a child’s lack of familiarity with the tasks possibly leads to a lower cognitive score in general and may also reflect only a developmental delay, but will not give sufficient information about the developmental potential. Additionally, a family’s low socioeconomic sta-tus could negatively influence the neurocognitive outcome, e.g., through less extensive support by parents, which could compensate for neurodevelopmental deficits. However, be-cause we did not assess socioeconomic status in our cohort study, this influence could not be explored.

ConclusionsNEL seems to be associated with similar neurodevel-

opmental and motor outcomes as are seen after other tech-niques of eliminating intraventricular blood components in neonatal posthemorrhagic hydrocephalus. Our results emphasize the relevance of using different neurodevel-opmental and motor outcome parameters for the evalua-tion of neurosurgical procedures in neonates, especially because the patient’s ability to walk freely seems to be more positively affected in comparison to BSID II MDI measurements. We describe the outcome after the NEL in neonatal posthemorrhagic hydrocephalus for the first time and, although our small sample size does not allow us to draw solid conclusions, it may, however, contribute to the ongoing debate of how to manage posthemorrhagic hydro-cephalus surgically. The neurodevelopmental outcome and brain volume should be further investigated in an extended setup in order to allow a comparison between different surgical procedures used for this patient cohort.

TABLE 4. Comparison of baseline characteristics between the DRIFT study and the current NEL study

Characteristic NEL Cohort DRIFT Cohort

No. of patients 42 34Median gestational age in wks (IQR)

27 (23–41) 27 (24–34)

Median birth weight in g (IQR) 1168 (520–3490) 1066 (640–2100)Male sex 64% 71%Parenchymal hemorrhagic infarction

50% 53%

Median age at intervention in days (IQR)

21 (8–77) 20 (7–28)

Based on data (DRIFT cohort) from Whitelaw et al.10



Behrens et al.

References 1. Badhiwala JH, Hong CJ, Nassiri F, et al. Treatment of post-

hemorrhagic ventricular dilation in preterm infants: a system-atic review and meta-analysis of outcomes and complications. J Neurosurg Pediatr. 2015; 16(5): 545–555.

2. Etus V, Kahilogullari G, Karabagli H, Unlu A. Early endo-scopic ventricular irrigation for the treatment of neonatal posthemorrhagic hydrocephalus: a feasible treatment option or not? A multicenter study. Turk Neurosurg. 2018; 28(1): 137–141.

3. Christian EA, Melamed EF, Peck E, et al. Surgical manage-ment of hydrocephalus secondary to intraventricular hemor-rhage in the preterm infant. J Neurosurg Pediatr. 2016; 17(3): 278–284.

4. d’Arcangues C, Schulz M, Bührer C, et al. Extended expe-rience with neuroendoscopic lavage for posthemorrhagic hydrocephalus in neonates. World Neurosurg. 2018; 116: e217–e224.

5. Papile LA, Burstein J, Burstein R, Koffler H. Incidence and evolution of subependymal and intraventricular hemorrhage: a study of infants with birth weights less than 1,500 gm. J Pediatr. 1978; 92(4): 529–534.

6. Cherian S, Whitelaw A, Thoresen M, Love S. The patho-genesis of neonatal post-hemorrhagic hydrocephalus. Brain Pathol. 2004; 14(3): 305–311.

7. Whitelaw A, Jary S, Kmita G, et al. Randomized trial of drainage, irrigation and fibrinolytic therapy for premature infants with posthemorrhagic ventricular dilatation: devel-opmental outcome at 2 years. Pediatrics. 2010; 125(4): e852–e858.

8. Gram M, Sveinsdottir S, Cinthio M, et al. Extracellular he-moglobin—mediator of inflammation and cell death in the choroid plexus following preterm intraventricular hemor-rhage. J Neuroinflammation. 2014; 11: 200.

9. Limbrick DD Jr, Mathur A, Johnston JM, et al. Neurosurgical treatment of progressive posthemorrhagic ventricular dilation in preterm infants: a 10-year single-institution study. J Neu-rosurg Pediatr. 2010; 6(3): 224–230.

10. Whitelaw A, Evans D, Carter M, et al. Randomized clinical trial of prevention of hydrocephalus after intraventricular hemorrhage in preterm infants: brain-washing versus tapping fluid. Pediatrics. 2007; 119(5): e1071–e1078.

11. Luyt K, Jary S, Lea C, et al. Ten-year follow-up of a ran-domised trial of drainage, irrigation and fibrinolytic therapy (DRIFT) in infants with post-haemorrhagic ventricular dila-tation. Health Technol Assess. 2019; 23(4): 1–116.

12. Schulz M, Bührer C, Pohl-Schickinger A, et al. Neuroendo-scopic lavage for the treatment of intraventricular hemor-rhage and hydrocephalus in neonates. J Neurosurg Pediatr. 2014; 13(6): 626–635.

13. Palisano R, Rosenbaum P, Bartlett D, et al. GMFCS E&R Gross Motor Function Classification System Expanded and Revised. CanChild Centre for Childhood Disability Research; 2007.

14. Palisano R, Rosenbaum P, Walter S, et al. Development and reliability of a system to classify gross motor function in children with cerebral palsy. Dev Med Child Neurol. 1997; 39(4): 214–223.

15. Yushkevich PA, Piven J, Hazlett HC, et al. User-guided 3D active contour segmentation of anatomical structures: signifi-cantly improved efficiency and reliability. Neuroimage. 2006; 31(3): 1116–1128.

16. Gelman A, Jakulin A, Pittau M, et al. A weakly informative default prior distribution for logistic and other regression models. Ann Appl Stat. 2009; 2(4): 1360–1383.

17. R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing; 2019.

18. Wickham H, Averick M, Bryan J, et al. Welcome to Tidy-verse. J Open Source Softw. 2019; 4(43): 1686.

19. Signorell A, Aho K, Alfons A, et al. DescTools: Tools for Descriptive Statistics. R package; 2020.

20. Gelman A, Hill J. Data Analysis Using Regression and Mul-tilevel/Hierarchical Models. Cambridge University Press; 2006

21. van Buuren S, Groothuis-Oudshoorn K. mice: Multivariate imputation by chained equations in R. J Stat Softw. 2011; 45(3): 1–67.

22. Cherukuri V, Ssenyonga P, Warf BC, et al. Learning based segmentation of CT brain images: application to postoper-ative hydrocephalic scans. IEEE Trans Biomed Eng. 2018; 65(8): 1871–1884.

23. Klimont M, Flieger M, Rzeszutek J, et al. Automated ven-tricular system segmentation in paediatric patients treated for hydrocephalus using deep learning methods. Biomed Res Int. 2019; 2019: 3059170.

24. Kulkarni AV, Schiff SJ, Mbabazi-Kabachelor E, et al. Endo-scopic treatment versus shunting for infant hydrocephalus in Uganda. N Engl J Med. 2017; 377(25): 2456–2464.

25. Mandell JG, Langelaan JW, Webb AG, Schiff SJ. Volumetric brain analysis in neurosurgery: Part 1. Particle filter seg-mentation of brain and cerebrospinal fluid growth dynamics from MRI and CT images. J Neurosurg Pediatr. 2015; 15(2): 113–124.

26. Brouwer MJ, de Vries LS, Kersbergen KJ, et al. Effects of posthemorrhagic ventricular dilatation in the preterm infant on brain volumes and white matter diffusion variables at term-equivalent age. J Pediatr. 2016; 168: 41–49.e1.

27. Cioni G, Bertuccelli B, Boldrini A, et al. Correlation between visual function, neurodevelopmental outcome, and magnetic resonance imaging findings in infants with periventricular leucomalacia. Arch Dis Child Fetal Neonatal Ed. 2000; 82(2): F134–F140.

28. Peterson M, Warf BC, Schiff SJ. Normative human brain volume growth. J Neurosurg Pediatr. 2018; 21(5): 478–485.

DisclosuresThe authors report no conflict of interest concerning the materi-als or methods used in this study or the findings specified in this paper.

Author ContributionsConception and design: Thomale, Tietze, Schulz. Acquisition of data: Thomale, Behrens, Tietze, Walch, Bittigau, Schulz. Analysis and interpretation of data: Thomale, Behrens, Tietze, Bührer, Schulz, Aigner. Drafting the article: Thomale, Behrens, Tietze, Bührer, Aigner. Critically revising the article: Thomale, Behrens, Tietze, Walch, Bittigau, Bührer, Aigner. Reviewed submitted version of manuscript: Thomale, Behrens, Tietze, Walch, Bittigau, Bührer, Aigner. Approved the final version of the manuscript on behalf of all authors: Thomale. Statistical analysis: Thomale, Bührer, Aigner. Administrative/technical/material support: Thomale, Aigner. Study supervision: Thomale, Aigner.

Supplemental InformationOnline-Only ContentSupplemental material is available with the online version of the article.

Supplemental Materials. https://thejns.org/doi/suppl/10.3171/ 2020.5.PEDS20211.

CorrespondenceUlrich-Wilhelm Thomale: Charité—Universitätsmedizin Berlin, Germany. [email protected].




Documents

Neurodevelopmental outcome at 2 years after