Preoperative risk and postoperative outcome from subdural fluid collections in African infants with postinfectious hydrocephalus

Jessica R. LaneDepartment of Neurosurgery, Penn State College of Medicine, Hershey;

Search for other papers by Jessica R. Lane in
jns
Google Scholar
PubMed
Close
 MD
,
Paddy SsentongoCenter for Neural Engineering, Department of Engineering Science and Mechanics, The Pennsylvania State University, University Park;
Department of Public Health Sciences, Penn State College of Medicine, Hershey;

Search for other papers by Paddy Ssentongo in
jns
Google Scholar
PubMed
Close
 MD, PhD, MPH
,
Mallory R. PetersonCenter for Neural Engineering, Department of Engineering Science and Mechanics, The Pennsylvania State University, University Park;

Search for other papers by Mallory R. Peterson in
jns
Google Scholar
PubMed
Close
 BS
,
Joshua R. HarperCenter for Neural Engineering, Department of Engineering Science and Mechanics, The Pennsylvania State University, University Park;

Search for other papers by Joshua R. Harper in
jns
Google Scholar
PubMed
Close
 PhD
,
Edith Mbabazi-KabachelorCURE Children’s Hospital of Uganda, Mbale;

Search for other papers by Edith Mbabazi-Kabachelor in
jns
Google Scholar
PubMed
Close
 MD, MSc
,
John MugambaCURE Children’s Hospital of Uganda, Mbale;

Search for other papers by John Mugamba in
jns
Google Scholar
PubMed
Close
 MD
,
Peter SsenyongaCURE Children’s Hospital of Uganda, Mbale;
Mulago National Referral Hospital, Kampala, Uganda;

Search for other papers by Peter Ssenyonga in
jns
Google Scholar
PubMed
Close
 MD
,
Justin OnenCURE Children’s Hospital of Uganda, Mbale;
Mulago National Referral Hospital, Kampala, Uganda;

Search for other papers by Justin Onen in
jns
Google Scholar
PubMed
Close
 MD
,
Ruth DonnellyDivision of Neurosurgery, University of Toronto, Hospital for Sick Children, Toronto;

Search for other papers by Ruth Donnelly in
jns
Google Scholar
PubMed
Close
 PhD
,
Jody LevenbachToronto Western Hospital, University Health Network, Toronto, Ontario, Canada; and

Search for other papers by Jody Levenbach in
jns
Google Scholar
PubMed
Close
 PhD
,
Venkateswararao CherukuriSchool of Electrical Engineering and Computer Science, The Pennsylvania State University, University Park;

Search for other papers by Venkateswararao Cherukuri in
jns
Google Scholar
PubMed
Close
 PhD
,
Vishal MongaSchool of Electrical Engineering and Computer Science, The Pennsylvania State University, University Park;

Search for other papers by Vishal Monga in
jns
Google Scholar
PubMed
Close
 PhD
,
Abhaya V. KulkarniDivision of Neurosurgery, University of Toronto, Hospital for Sick Children, Toronto;

Search for other papers by Abhaya V. Kulkarni in
jns
Google Scholar
PubMed
Close
 MD, PhD
,
Benjamin C. WarfDepartment of Neurosurgery, Boston Children’s Hospital and Department of Global Health and Social Medicine, Harvard Medical School, Boston, Massachusetts

Search for other papers by Benjamin C. Warf in
jns
Google Scholar
PubMed
Close
 MD
, and
Steven J. SchiffDepartment of Neurosurgery, Penn State College of Medicine, Hershey;
Center for Neural Engineering, Department of Engineering Science and Mechanics, The Pennsylvania State University, University Park;
Department of Physics, The Pennsylvania State University, University Park, Pennsylvania;

Search for other papers by Steven J. Schiff in
jns
Google Scholar
PubMed
Close
 MD, PhD
Free access

OBJECTIVE

This study investigated the incidence of postoperative subdural collections in a cohort of African infants with postinfectious hydrocephalus. The authors sought to identify preoperative factors associated with increased risk of development of subdural collections and to characterize associations between subdural collections and postoperative outcomes.

METHODS

The study was a post hoc analysis of a randomized controlled trial at a single center in Mbale, Uganda, involving infants (age < 180 days) with postinfectious hydrocephalus randomized to receive either an endoscopic third ventriculostomy plus choroid plexus cauterization or a ventriculoperitoneal shunt. Patients underwent assessment with the Bayley Scales of Infant and Toddler Development, Third Edition (Bayley-III; sometimes referred to as BSID-III) and CT scans preoperatively and then at 6, 12, and 24 months postoperatively. Volumes of brain, CSF, and subdural fluid were calculated, and z-scores from the median were determined from normative curves for CSF accumulation and brain growth. Linear and logistic regression models were used to characterize the association between preoperative CSF volume and the postoperative presence and size of subdural collection 6 and 12 months after surgery. Linear regression and smoothing spline ANOVA were used to describe the relationship between subdural fluid volume and cognitive scores. Causal mediation analysis distinguished between the direct and indirect effects of the presence of a subdural collection on cognitive scores.

RESULTS

Subdural collections were more common in shunt-treated patients and those with larger preoperative CSF volumes. Subdural fluid volumes were linearly related to preoperative CSF volumes. In terms of outcomes, the Bayley-III cognitive score was linearly related to subdural fluid volume. The distribution of cognitive scores was significantly different for patients with and those without subdural collections from 11 to 24 months of age. The presence of a subdural collection was associated with lower cognitive scores and smaller brain volume 12 months after surgery. Causal mediation analysis demonstrated evidence supporting both a direct (76%) and indirect (24%) effect (through brain volume) of subdural collections on cognitive scores.

CONCLUSIONS

Larger preoperative CSF volume and shunt surgery were found to be risk factors for postoperative subdural collection. The size and presence of a subdural collection were negatively associated with cognitive outcomes and brain volume 12 months after surgery. These results have suggested that preoperative CSF volumes could be used for risk stratification for treatment decision-making and that future clinical trials of alternative shunt technologies to reduce overdrainage should be considered.

ABBREVIATIONS

Bayley-III = Bayley Scales of Infant and Toddler Development, Third Edition; ETV/CPC = endoscopic third ventriculostomy plus choroid plexus cauterization; ITT = intention to treat; PIH = postinfectious hydrocephalus; TR = treatment received; VPS = ventriculoperitoneal shunt.

OBJECTIVE

This study investigated the incidence of postoperative subdural collections in a cohort of African infants with postinfectious hydrocephalus. The authors sought to identify preoperative factors associated with increased risk of development of subdural collections and to characterize associations between subdural collections and postoperative outcomes.

METHODS

The study was a post hoc analysis of a randomized controlled trial at a single center in Mbale, Uganda, involving infants (age < 180 days) with postinfectious hydrocephalus randomized to receive either an endoscopic third ventriculostomy plus choroid plexus cauterization or a ventriculoperitoneal shunt. Patients underwent assessment with the Bayley Scales of Infant and Toddler Development, Third Edition (Bayley-III; sometimes referred to as BSID-III) and CT scans preoperatively and then at 6, 12, and 24 months postoperatively. Volumes of brain, CSF, and subdural fluid were calculated, and z-scores from the median were determined from normative curves for CSF accumulation and brain growth. Linear and logistic regression models were used to characterize the association between preoperative CSF volume and the postoperative presence and size of subdural collection 6 and 12 months after surgery. Linear regression and smoothing spline ANOVA were used to describe the relationship between subdural fluid volume and cognitive scores. Causal mediation analysis distinguished between the direct and indirect effects of the presence of a subdural collection on cognitive scores.

RESULTS

Subdural collections were more common in shunt-treated patients and those with larger preoperative CSF volumes. Subdural fluid volumes were linearly related to preoperative CSF volumes. In terms of outcomes, the Bayley-III cognitive score was linearly related to subdural fluid volume. The distribution of cognitive scores was significantly different for patients with and those without subdural collections from 11 to 24 months of age. The presence of a subdural collection was associated with lower cognitive scores and smaller brain volume 12 months after surgery. Causal mediation analysis demonstrated evidence supporting both a direct (76%) and indirect (24%) effect (through brain volume) of subdural collections on cognitive scores.

CONCLUSIONS

Larger preoperative CSF volume and shunt surgery were found to be risk factors for postoperative subdural collection. The size and presence of a subdural collection were negatively associated with cognitive outcomes and brain volume 12 months after surgery. These results have suggested that preoperative CSF volumes could be used for risk stratification for treatment decision-making and that future clinical trials of alternative shunt technologies to reduce overdrainage should be considered.

In Brief

This study characterized subdural collections in African infants with postinfectious hydrocephalus following surgical treatment. Using normalized volumetrics for brain and CSF, higher preoperative CSF volume was associated with both higher rates and larger volumes of subdural accumulation. The presence and size of the subdural collection were associated with lower cognitive scores. These results suggest that preoperative CSF volumes could be used for risk stratification for treatment decision-making and for future clinical trials of alternative shunt technologies.

The development of subdural hematomas and hygromas has been a long-noted complication of CSF diversion in hydrocephalus.1,2 Series in the literature have reported postoperative incidence rates of 1% to 3%; however, many of these studies did not include scheduled imaging for patients in the absence of concerning symptoms.36 The progress in tools used to calculate the volume of brain, CSF, and subdural collections from pre- and postoperative imaging allows us to use these volumes as additional clinical predictors and outcome variables.7,8 The analysis of the volume of intracranial components has been recently aided by the establishment of normal growth curves with which volumes in patients with hydrocephalus may be standardized to age- and sex-adjusted normative values.9 However, to our knowledge, normative brain growth data have not been used to evaluate the impact of subdural fluid collections in infant hydrocephalus. The ability to quantify preoperative intracranial contents in this way allows us to enhance the prediction of patients at the highest subdural risk. Such preoperative assessments can offer alternative treatment management strategies to reduce subdural accumulations affecting cognitive outcome, brain growth, and the risk of further surgical intervention.

This study describes the development of postoperative subdural collections in a population of infants with hydrocephalus in sub-Saharan Africa. These patients are part of an ongoing randomized controlled trial at a single center in Mbale, Uganda, involving infants with postinfectious hydrocephalus (PIH) randomized to receive either endoscopic third ventriculostomy plus choroid plexus cauterization (ETV/CPC) or a ventriculoperitoneal shunt (VPS).10 To assess surgical outcomes, these patients underwent longitudinal cognitive testing as well as serial, scheduled imaging.

Methods

Patients

This study analyzes the imaging and outcomes from a randomized controlled trial of ETV/CPC versus VPS for PIH in Ugandan infants (ClinicalTrials.gov registration no. NCT01936272). Patients were included if they were younger than 180 days of age at the time of the randomization, met criteria for PIH, and were judged by two independent neurosurgeons to be appropriate candidates for either ETV/CPC or a VPS. Patients were assigned to a surgical treatment randomly in a nonstratified 1:1 scheme. ETV/CPC was performed with a flexible endoscope, with an opening made in the floor of the third ventricle to allow passage of CSF into the subarachnoid cisterns, followed by the bilateral cauterization of the choroid plexus from the foramen of Monro to the anterior temporal horn. For VPS surgeries, a Chhabra shunt (Surgiwear) was used. In patients randomized to ETV/CPC, if the floor of the third ventricle could not be opened or there was intraoperative evidence of scarring in the prepontine cistern, the surgeon placed a VPS. Nine patients randomized to ETV/CPC had a VPS placed at the initial surgery. We report both the intention to treat (ITT) and the treatment received (TR) in our results to account for this crossover. Patients underwent blinded neuropsychological assessment using the Bayley Scales of Infant and Toddler Development, Third Edition (Bayley-III; sometimes referred to as BSID-III)21 and CT imaging preoperatively and again at 6, 12, and 24 months postoperatively. The full details of the clinical trial design, enrollment, and treatments are discussed in the publication of the year 110 and year 2 results.11

Imaging Analysis

We used a previously published algorithm for semiautomated segmentation of CT scans to calculate the volume of brain, CSF, and subdural fluid.8 Images were reviewed for initial segmentation accuracy by two authors (J.R.L. and S.J.S.) and manually revised as necessary by the same two authors. Extraaxial fluid collections were segmented as subdural fluid regardless of density or size, including subdural hematomas of varying chronicity and subdural hygromas. Segmentation of these collections was not carried forward into the region of the sylvian, central, or interhemispheric fissures. The majority of subdural collections were located in the cortical convexity region, with some extending over the tentorium. We calculated age- and sex-adjusted z-scores from the median brain and CSF volumes using normative curves for brain growth and CSF accumulation derived from the NIH Pediatric MRI Data Repository.9

Statistical Analysis

Demographics of the patients with and without subdural collections were compared using the Fisher’s exact and Wilcoxon rank-sum tests. We examined the association of preoperative CSF and brain volume z-scores with subsequent subdural volume using linear regression. Logistic regression models were created to assess the likelihood of subdural incidence based on preoperative CSF volume z-score. Separate models were created for both surgery types, defined as ITT or as TR. Univariate and multivariable logistic regression analyses were then used to evaluate the effect of the surgery received and preoperative CSF volume z-score on the development of a subdural collection in all patients at the 6- and 12-month time points, and odds ratios with 95% confidence intervals were generated.

We examined the association between subdural volume and Bayley-III cognitive score with linear regression. Nonparametric smoothing spline ANOVA was used to examine the longitudinal differences in the cognitive score over patient ages in those with and without subdural collections. The smoothing splines were calculated by minimizing a penalized least-square functional, which included smoothing parameters determined by an iterative cross-validation process.12 Cognitive scores were modeled using age and subdural presence as main factors, with an interaction term and subject-specific random effects. We summarized the effect of subdural presence on outcomes 12 months after surgery using the Mann-Whitney U-test with Hodges-Lehmann estimates of confidence intervals for cognitive scores and brain volume z-scores. Overall patient survival and failure-free survival curves were assessed using the Kaplan-Meier method with the log-rank test.

To examine whether brain volume mediates the association between subdural collection and cognitive outcome, a causal mediation analysis was performed.

Mediation is a hypothesized causal chain in which the effect of an independent variable, X, on a dependent variable, Y, is transmitted through a third intervening or mediating variable, M.13 In other words, X affects M, which in turn affects Y. Mediation can be depicted when the effect of X on Y is through M, as X → M → Y. The mediation effect in which X → M → Y is called the indirect effect. The effect of X on Y that does not pass through M, X → Y, is called the direct effect.

Our empirical goal is to separate the total effects observed in the Bayley-III scores into the direct effects acting through the subdural collection and the indirect effects through the mediator of brain volume. We use the standard mediation model that includes multiple ordinary-least-squares regression models, which take the form: M = β0 + Xβ1 + ε and Y = θ0 + Xθ1 + Mθ2 + δ, where M is the mediating variable (brain volume), X is a vector of explanatory (subdural collection) and confounding variables (age, sex, baseline Bayley-III score), β1 and θ1 are vectors of parameters to be estimated, Y is the follow-up Bayley-III score, and ε and δ are error terms. In essence, causal mediation analysis inserts a regression for the mediation, M, into the regression that includes the rest of the variables and factors.

Inserting the linear model for M into the linear model for Y yields Y = (θ0 + θ2β0) + X(θ1 + θ2β1) + (θ2ε + δ). θ1 and θ2β1 represent the direct and indirect effects, respectively, of the exposure on the outcome.

Using the open-source mediation package in R (The R Project),14 we performed a causal mediation analysis with 1000 Monte Carlo draws to estimate confidence intervals. Statistical analyses were performed using R version 3.6.2 (R Core Team) and figures were plotted using MATLAB version 2020b (The MathWorks, Inc.).

Results

Patients

Overall, 100 patients from the randomized controlled trial were included in the analysis. Of these patients, 48 had a subdural collection detectable on at least one postoperative CT scan. In the shunt-treatment arm of the trial, 67.4% (33 of 49) of patients by ITT and 66.7% (38 of 57) of patients by TR developed subdural collections over the course of the 24 months after surgery. When we restricted to patients who did not require surgical revision prior to the 6-month scan, subdural collections were seen at 6 months in 51.7% (17 of 33) of patients by ITT and 55.3% (21 of 38) of patients by TR in the shunt-treatment arm of the study (Table 1). The patients who had developed subdural collections 6 months after surgery were significantly older, had larger preoperative CSF volumes, and were more likely to have been in the VPS arm of the study, but there was no difference in preoperative cognitive scores or brain volume (Table 2). The crossover from ETV/CPC to a VPS resulted in a different number of patients initially randomized to a VPS (n = 23) versus the number of patients who actually received a VPS (n = 27) (Table 2). Subdural collection volumes 6 months after surgery ranged in size from < 50 cm3 to > 500 cm3, with the majority being < 150 cm3 (Fig. 1). Subdural volumes were not static, and they changed over the course of the study, some appearing at later time points than the first scheduled 6-month CT scan.

TABLE 1.

Rates of subdural collections

Grouping MethodOpNo. of PatientsSubdural Collections% Affected
6-mo postop, no revisions
 ITTETV/CPC29620.7%
VPS331751.5%
 TRETV/CPC2428.3%
VPS382155.3%
 Total622337.1%
All time points & patients
 ITTETV/CPC511529.4%
VPS493367.4%
 TRETV/CPC431023.3%
VPS573866.7%
 Total1004848.0%
TABLE 2.

Demographics of patients with and those without subdural collections 6 months postoperatively

CharacteristicSubdural Collections 6 Mos Postop (n = 33)No Subdural Collections 6 Mos Postop (n = 54)
Female sex, n (%)9 (27)25 (46)
Median age preop, days (IQR)*107 (86–147)90 (61.5–110.25)
Median preop brain vol, cm3 (IQR) 445 (375.25–525.75)461.5 (379.75–499.25)
Median preop tissue z-score (IQR)−3.56 (−4.62 to −1.94)−3.08 (−4.11 to −1.64)
Median preop CSF vol, cm3 (IQR)*936 (679.5–1365.5)751.5 (502.25–929.75)
Median preop CSF z-score (IQR)*7.91 (6.58–8.30)7.00 (5.73–8.00)
Median preop Bayley-III score (IQR)1 (1–2)2 (1–5)
Randomized to VPS, n (%)*23 (70)21 (39)
Received VPS, n (%)*27 (82)24 (44)

Only those patients with 6-month postoperative imaging are included.

Significant at p < 0.05, Fisher’s exact or Wilcoxon rank-sum test.

FIG. 1
FIG. 1

Bar graph showing subdural volumes 6 months after surgery with CT scans showing examples of the segmentation of brain (green), CSF (red), and subdural fluid collection (blue). The example segmentation images are selected from the smallest and largest volume bins in the histogram. Figure is available in color online only.

Preoperative Predictors of Subdural Incidence

For all patients, there was a significant linear relationship between preoperative CSF volume z-score and subdural volume 6 months after surgery (p = 0.003; Fig. 2A). In Fig. 2, we visually designated ETV/CPC and VPS by TR, but the regressions are calculated independent of the assignment to a treatment arm. To be cautious for potential confounding bias, we compared and found no preoperative difference in CSF accumulation (by volume or z-scores) in either ITT or TR, or even per protocol (eliminating crossovers); these comparisons are shown in Supplementary Fig. 1. This linear association remained significant when analysis was limited to nonzero subdural volumes (p = 0.017; Fig. 2C). There were no associations between preoperative brain volume and subdural volume 6 months after surgery (Fig. 2B and D).

FIG. 2
FIG. 2

Scatterplots. A and B: Subdural volume for all patients 6 months after surgery. C and D: Preoperative CSF and brain volumes for patients with nonzero subdural collections present 6 months after surgery. The linear regression reaches statistical significance using preoperative CSF z-score (A and C) but not preoperative brain volume z-score (B and D). Adjusted R2 values are 0.087 (A), 0.020 (B), 0.154 (C), and 0.032 (D). Figure is available in color online only.

In the VPS arm, preoperative CSF volume z-scores were predictive of subdural presence 6 months after surgery by TR (logistic regression, p = 0.047; Fig. 3 top row) and 12 months after surgery by both ITT and TR (p = 0.038 and p = 0.036, respectively; Fig. 3 middle row). This predictability dissipated by 24 months postsurgery (Fig. 3 bottom row). In the ETV/CPC arm, preoperative CSF volume z-scores were not predictive of subdural presence at any time point by either ITT or TR (Fig. 3).

FIG. 3
FIG. 3

Likelihood of subdural development. The six panels represent logistic regression for developing a subdural collection at each of the three specified time points (6, 12, and 24 months postoperatively) for patients in the ETV/CPC or VPS arms of the study, as specified by either ITT or TR. The regressions reach significance for patients in the VPS arm only at the 6- and 12-month time points by TR, and at the 6-month time point by ITT. Figure is available in color online only.

We generated a relative odds ratio of the risk of subdural presence using a multivariable logistic regression that included both preoperative CSF volume z-score and ETV/CPC versus a VPS by TR. Compared with ETV/CPC, a VPS was associated with a higher relative risk of subdural presence with ORs of 6.53 (95% CI 2.27–22.07; p = 0.001) 6 months after surgery and 6.10 (95% CI 2.26–18.76; p = 0.0007) 12 months after surgery. Each unit increase in preoperative CSF volume z-score was associated with a 62% or 56% increase in the odds of subdural presence at 6 or 12 months after surgery (p = 0.03 and p = 0.04, respectively; Table 3).

TABLE 3.

Predictors of subdural collections at the 6- and 12-month time points

PredictorUnivariate Logistic RegressionMultivariable Logistic Regression
OR (95% CI)p ValueOR (95% CI)p Value
6-mo
 TR
  VPS5.63 (2.10–17.10)0.0016.53 (2.27–22.07)0.001
  ETV/CPCReference
 Preop CSF vol z-scores1.67 (1.12–2.63)0.021.62 (1.07–2.58)0.03
12-mo
 TR
  VPS6.35 (2.41–19.09)0.00046.10 (2.26–18.76)0.0007
  ETV/CPCReference
 Preop CSF vol z-scores1.58 (1.08–2.43)0.031.56 (1.04–2.44)0.04

Association Between Subdural Collections and 12-Month Outcomes

Compared with patients without subdural collections, those with subdural collections had a lower median Bayley-III cognitive score and lower brain volume z-scores 12 months after surgery (p = 0.005 and p = 0.04, respectively; Fig. 4A and B). Cognitive scores at 12 months after surgery had a significant linear relationship with subdural volume at the same time point (p = 0.023; Fig. 4C). Longitudinal analysis of cognitive scores in patients with and those without subdural collections showed a difference between the two groups from 11 to 24 months of age (Fig. 4D). Overall patient mortality and surgical failure-free survival were not different between the patients with and without subdural collections (p = 0.45 and p = 0.13 respectively; Supplementary Fig. 2).

FIG. 4
FIG. 4

Subdural collections and outcomes 12 months after surgery. A: Boxplot showing that Bayley-III cognitive scores in the subdural presence and absence groups are significantly different by the Wilcoxon rank-sum test (p = 0.005). B: Boxplot demonstrating that brain volume z-scores in the subdural presence and absence groups are significantly different by the Wilcoxon rank-sum test (p = 0.04). C: Scatterplot showing that the linear regression model of subdural volume and cognitive score 12 months after surgery reaches significance with p = 0.023 and an adjusted R2 of 0.047. D: Scatterplot showing the Bayley-III cognitive score smoothing spline ANOVA (SSANOVA) fit and 95% confidence intervals for the subdural presence and absence groups across all patient ages. The gray highlighted region, where the confidence intervals do not overlap, shows the age range for which there is a significant difference between the two groups with regard to cognitive score. Figure is available in color online only.

The effect on cognitive outcome could be a direct effect of the subdural collection on brain function (e.g., via deformation, membrane formation, or inflammation15) or might be mediated through an effect on brain growth over time as measured in age-adjusted brain volume. Such direct and indirect effects can be analyzed with a causal mediation analysis. The total effect of subdural presence on Bayley-III cognitive score was estimated at −2.0 (95% CI −3.46 to −0.57; p = 0.006). Of this, 24% was mediated by brain volume (p = 0.044). The remaining 76% could be attributed to a direct effect of the subdural collection on the cognitive score (p = 0.036; Fig. 5).

FIG. 5
FIG. 5

Causal mediation analysis. The causal mediation model separates the direct and indirect effects of subdural collections on cognitive outcomes as illustrated in the schematic. The table shows the effect sizes for total, direct, and indirect effects as well as the proportion mediated (24% indirect and 76% direct). The 95% confidence intervals and p values are also included, with all values reaching statistical significance with p < 0.05. Figure is available in color online only.

Discussion

Patients in this study, especially those with larger preoperative CSF volumes who underwent VPS treatment, had high rates of developing postoperative subdural collections. Patients with subdural collections had worse cognitive outcomes and lower brain volumes, although mortality was similar.

This randomized study is especially well suited to capture the characteristics of subdural collections due to repeated, scheduled imaging that permitted the detection of even asymptomatic accumulations. Additionally, the patient characteristics in this trial are characterized by their postinfectious etiology, the time elapsed prior to presentation and treatment, the specific surgical protocol, and the African location.

The high frequency of subdural collections in this patient population enabled us to explore associations with both preoperative factors and postoperative outcomes. In previous research, the risk of developing a subdural collection has been ascribed to brain atrophy and a thin cortical mantle,16,17 but we did not find a relationship between preoperative age- and sex-normalized brain volume and development of a subdural collection. In contrast, we found that preoperative CSF volume was associated with both higher rates and larger volumes of subdural collections. These results suggest that preoperative CSF volumes could be used for preoperative risk stratification for treatment decision-making. The older age and higher preoperative CSF volumes of the children who developed subdural collections imply that delay to treatment was a possible significant risk factor, but we are unable to identify the date of onset of hydrocephalus for all patients in the cohort. Patients with subdural accumulations had preoperative brain volumes and cognitive scores that were equivalent to patients who did not develop subdural collections. We did not identify features attributable to the initial infection that were associated with the difference in cognitive outcome other than increased CSF preoperatively and the presence of subdural collections postoperatively.

Subdural presence and volume were both related to cognitive scores. Nevertheless, we did not find a clear threshold volume discriminating between consequential and harmless subdural collections—the presence of a subdural collection itself appeared to place a patient into a higher-risk outcome category. As we analyzed brain volume compared with age- and sex-adjusted normative values, a decreased brain volume z-score could reflect either reduced brain growth11 or a true loss of volume, as seen with chronic subdural hematomas in elderly patients with increased brain atrophy.18 Our causal mediation analysis was consistent with the subdural collections having both a direct effect on cognitive function and an effect mediated through brain growth and volume.

Mortality was not greater in patients with subdural collections, which was unexpected given the significant excess mortality associated with chronic subdural hematomas in the elderly.19 It is possible that the 15% mortality rate at 2 years in the current study11 obscured a relationship between observed subdural collections and survivability.

There are a number of limitations to this study. Since subdural accumulations of this frequency were not anticipated in the clinical trial design, the present analysis represents a post hoc analysis of findings that were not primary or secondary outcome measures. The study did not plan for routine imaging before 6 months postsurgery in children with uncomplicated postoperative courses, and any effect of rapid expansion of the cerebral mantle following decompression, contributing to an increase in brain volume, could not be isolated from brain growth. In addition, we could not assess brain compliance that could have affected the mechanical resistance to subdural accumulation. Since this study was focused on a single etiology of hydrocephalus and a single clinical treatment site, the generalizability of these findings is limited given the nature and severity of the PIH in this cohort, the time elapsed prior to presentation and treatment, and the specific surgical protocol and shunt technology employed.

As a clinical trial in an African country with limited resources, the trial protocol incorporated the customary inexpensive shunt (Chhabra) in standard use at the surgical site and other sites in resource-limited settings. Prior study of the performance of this shunt had been shown to be comparable to a more expensive shunt.20 Nevertheless, the results of the current study suggest that investigation of alternative shunt technology, addressing differential valve pressure or antisiphon characteristics, would be timely to consider. It is intriguing to speculate whether reducing the frequency and size of subdural collections through alternative shunt technology might have affected the clinical trial primary cognitive outcome.

Conclusions

This study characterizes subdural collections in African infants with PIH treated with ETV/CPC or a VPS. Higher preoperative CSF volume was associated with both higher rates and larger volumes of subdural accumulation. The presence and size of subdural collections were associated with lower cognitive scores. This effect appears to be related to a direct effect of the subdural collection on the brain, and to be mediated in part through a reduction in brain volume. These results suggest that preoperative CSF volumes could be used for risk stratification for treatment decision-making, and that future clinical trials of alternative shunt technologies should be considered.

Acknowledgments

Supported by the National Institutes of Health grants R01HD085853 and 1F30HD102120.

Appendix

Steering Committee

Benjamin C. Warf (Harvard University, USA), Steven J. Schiff (Pennsylvania State University, USA), and Abhaya V. Kulkarni (University of Toronto, Canada).

Investigators

Edith Mbabazi (CURE Children’s Hospital of Uganda, Uganda), John Mugamba (CURE Children’s Hospital of Uganda, Uganda), Peter Ssenyonga (CURE Children’s Hospital of Uganda, Uganda), Justin Onen (CURE Children’s Hospital of Uganda, Uganda), Ruth Donnelly (The Hospital for Sick Children, Canada), Jody Levenbach (The Hospital for Sick Children, Canada), Vishal Monga (Pennsylvania State University, USA), Mallory Peterson (Pennsylvania State University, USA), Venkateswararao Cherukuri (Pennsylvania State University, USA), Jessica Lane (Pennsylvania State University, USA), Paddy Ssentongo (Pennsylvania State University, USA), and Joshua Harper (Pennsylvania State University, USA).

Data Safety Monitoring Board

Jay Riva-Cambrin (Chair) (University of Calgary, Canada), Michael Scott (Harvard University, USA), Graham Fieggen (University of Cape Town, South Africa), Julius Kiwanuka (Mbarara University of Science & Technology, Uganda), and Francis Bajunirwe (Mbarara University of Science & Technology, Uganda).

Research Assistants

Esther Nalule, Julie Johnson, John Kimbugwe, and Brian Nsubuga Kaaya.

Disclosures

The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper.

Author Contributions

Conception and design: Schiff, Lane, Kulkarni, Warf. Acquisition of data: Mbabazi-Kabachelor, Mugamba, Ssenyonga, Onen. Analysis and interpretation of data: Schiff, Lane, Ssentongo, Peterson, Harper, Mbabazi-Kabachelor, Donnelly, Levenbach, Cherukuri, Monga, Kulkarni. Drafting the article: Schiff, Lane, Ssentongo, Peterson, Harper. Critically revising the article: Schiff, Lane, Ssentongo, Peterson, Harper, Kulkarni, Warf. Reviewed submitted version of manuscript: all authors. Approved the final version of the manuscript on behalf of all authors: Schiff. Statistical analysis: Schiff, Lane, Ssentongo, Peterson, Harper. Study supervision: Schiff, Kulkarni, Warf.

Supplemental Information

Online-Only Content

Supplemental material is available with the online version of the article.

References

  • 1

    Foltz EL, Shurtleff DB. Five-year comparative study of hydrocephalus in children with and without operation (113 cases). J Neurosurg. 1963;20:10641079.

  • 2

    Hoppe-Hirsch E, Sainte Rose C, Renier D, Hirsch JF. Pericerebral collections after shunting. Childs Nerv Syst. 1987;3(2):97102.

  • 3

    Lee JY, Wang KC, Cho BK. Functioning periods and complications of 246 cerebrospinal fluid shunting procedures in 208 children. J Korean Med Sci. 1995;10(4):275280.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 4

    Martínez-Lage JF, Pérez-Espejo MA, Almagro MJ, Ros de San Pedro J, López F, Piqueras C, Tortosa J. Syndromes of overdrainage of ventricular shunting in childhood hydrocephalus. Article in Spanish. Neurocirugia (Astur). 2005;16(2):124133.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 5

    Garton HJ, Kestle JR, Drake JM. Predicting shunt failure on the basis of clinical symptoms and signs in children. J Neurosurg. 2001;94(2):202210.

  • 6

    Zemack G, Romner B. Seven years of clinical experience with the programmable Codman Hakim valve: a retrospective study of 583 patients. J Neurosurg. 2000;92(6):941948.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 7

    Mandell JG, Kulkarni AV, Warf BC, Schiff SJ. Volumetric brain analysis in neurosurgery: Part 2. Brain and CSF volumes discriminate neurocognitive outcomes in hydrocephalus. J Neurosurg Pediatr. 2015;15(2):125132.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 8

    Cherukuri V, Ssenyonga P, Warf BC, Kulkarni AV, Monga V, Schiff SJ. Learning based segmentation of CT brain images: application to postoperative hydrocephalic scans. IEEE Trans Biomed Eng. 2018;65(8):18711884.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 9

    Peterson MR, Cherukuri V, Paulson JN, Ssentongo P, Kulkarni AV, Warf BC, et al. Normal childhood brain growth and a universal sex and anthropomorphic relationship to cerebrospinal fluid. J Neurosurg Pediatr. Published online July 9, 2021.doi:10.3171/2021.2.PEDS201006

    • Search Google Scholar
    • Export Citation
  • 10

    Kulkarni AV, Schiff SJ, Mbabazi-Kabachelor E, Mugamba J, Ssenyonga P, Donnelly R, et al. Endoscopic treatment versus shunting for infant hydrocephalus in Uganda. N Engl J Med. 2017;377(25):24562464.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 11

    Schiff S, Kulkarni A, Mbabazi-Kabachelor E, Mugamba J, Ssenyonga P, Donnelly R, et al. Brain growth after surgical treatment of infant post-infectious hydrocephalus in sub-Saharan Africa: two-year results of a randomized trial. J Neurosurg Pediatr. Published online July 9, 2021.doi:10.3171/2021.2.PEDS20949

    • Search Google Scholar
    • Export Citation
  • 12

    Gu C. Smoothing spline ANOVA models. Springer Series in Statistics.Springer;2002.

  • 13

    Pearl J, Mackenzie D. The Book of Why: The New Science of Cause and Effect. 1st ed. Basic Books;2018.

  • 14

    Tingley D, Yamamoto T, Hirose K, Keele L, Imai K. mediation: R package for causal mediation analysis. J Stat Softw. 2014;59(5).doi:10.18637/jss.v059.i05

    • Search Google Scholar
    • Export Citation
  • 15

    Won SY, Konczalla J, Dubinski D, Cattani A, Cuca C, Seifert V, et al. A systematic review of epileptic seizures in adults with subdural haematomas. Seizure. 2017;45:2835.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 16

    Puca A, Fernandez E, Colosimo C, Lauretti L, Pallini R, Tamburrini G. Hydrocephalus and macrocrania: surgical or non-surgical treatment of postshunting subdural hematoma?. Surg Neurol. 1996;45(4):376382.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 17

    Aguiar PH, Shu EB, Freitas AB, Leme RJ, Miura FK, Marino R Jr. Causes and treatment of intracranial haemorrhage complicating shunting for paediatric hydrocephalus. Childs Nerv Syst. 2000;16(4):218221.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 18

    Bin Zahid A, Balser D, Thomas R, Mahan MY, Hubbard ME, Samadani U. Increase in brain atrophy after subdural hematoma to rates greater than associated with dementia. J Neurosurg. 2018;129(6):15791587.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 19

    Miranda LB, Braxton E, Hobbs J, Quigley MR. Chronic subdural hematoma in the elderly: not a benign disease. J Neurosurg. 2011;114(1):7276.

  • 20

    Warf BC. Comparison of 1-year outcomes for the Chhabra and Codman-Hakim Micro Precision shunt systems in Uganda: a prospective study in 195 children. J Neurosurg. 2005;102(4 suppl):358362.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 21

    Bayley N. Bayley Scales of Infant and Toddler Development,. Third Edition. Pearson;2005.

  • Collapse
  • Expand
  • View in gallery
    FIG. 1

    Bar graph showing subdural volumes 6 months after surgery with CT scans showing examples of the segmentation of brain (green), CSF (red), and subdural fluid collection (blue). The example segmentation images are selected from the smallest and largest volume bins in the histogram. Figure is available in color online only.

  • View in gallery
    FIG. 2

    Scatterplots. A and B: Subdural volume for all patients 6 months after surgery. C and D: Preoperative CSF and brain volumes for patients with nonzero subdural collections present 6 months after surgery. The linear regression reaches statistical significance using preoperative CSF z-score (A and C) but not preoperative brain volume z-score (B and D). Adjusted R2 values are 0.087 (A), 0.020 (B), 0.154 (C), and 0.032 (D). Figure is available in color online only.

  • View in gallery
    FIG. 3

    Likelihood of subdural development. The six panels represent logistic regression for developing a subdural collection at each of the three specified time points (6, 12, and 24 months postoperatively) for patients in the ETV/CPC or VPS arms of the study, as specified by either ITT or TR. The regressions reach significance for patients in the VPS arm only at the 6- and 12-month time points by TR, and at the 6-month time point by ITT. Figure is available in color online only.

  • View in gallery
    FIG. 4

    Subdural collections and outcomes 12 months after surgery. A: Boxplot showing that Bayley-III cognitive scores in the subdural presence and absence groups are significantly different by the Wilcoxon rank-sum test (p = 0.005). B: Boxplot demonstrating that brain volume z-scores in the subdural presence and absence groups are significantly different by the Wilcoxon rank-sum test (p = 0.04). C: Scatterplot showing that the linear regression model of subdural volume and cognitive score 12 months after surgery reaches significance with p = 0.023 and an adjusted R2 of 0.047. D: Scatterplot showing the Bayley-III cognitive score smoothing spline ANOVA (SSANOVA) fit and 95% confidence intervals for the subdural presence and absence groups across all patient ages. The gray highlighted region, where the confidence intervals do not overlap, shows the age range for which there is a significant difference between the two groups with regard to cognitive score. Figure is available in color online only.

  • View in gallery
    FIG. 5

    Causal mediation analysis. The causal mediation model separates the direct and indirect effects of subdural collections on cognitive outcomes as illustrated in the schematic. The table shows the effect sizes for total, direct, and indirect effects as well as the proportion mediated (24% indirect and 76% direct). The 95% confidence intervals and p values are also included, with all values reaching statistical significance with p < 0.05. Figure is available in color online only.

  • 1

    Foltz EL, Shurtleff DB. Five-year comparative study of hydrocephalus in children with and without operation (113 cases). J Neurosurg. 1963;20:10641079.

  • 2

    Hoppe-Hirsch E, Sainte Rose C, Renier D, Hirsch JF. Pericerebral collections after shunting. Childs Nerv Syst. 1987;3(2):97102.

  • 3

    Lee JY, Wang KC, Cho BK. Functioning periods and complications of 246 cerebrospinal fluid shunting procedures in 208 children. J Korean Med Sci. 1995;10(4):275280.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 4

    Martínez-Lage JF, Pérez-Espejo MA, Almagro MJ, Ros de San Pedro J, López F, Piqueras C, Tortosa J. Syndromes of overdrainage of ventricular shunting in childhood hydrocephalus. Article in Spanish. Neurocirugia (Astur). 2005;16(2):124133.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 5

    Garton HJ, Kestle JR, Drake JM. Predicting shunt failure on the basis of clinical symptoms and signs in children. J Neurosurg. 2001;94(2):202210.

  • 6

    Zemack G, Romner B. Seven years of clinical experience with the programmable Codman Hakim valve: a retrospective study of 583 patients. J Neurosurg. 2000;92(6):941948.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 7

    Mandell JG, Kulkarni AV, Warf BC, Schiff SJ. Volumetric brain analysis in neurosurgery: Part 2. Brain and CSF volumes discriminate neurocognitive outcomes in hydrocephalus. J Neurosurg Pediatr. 2015;15(2):125132.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 8

    Cherukuri V, Ssenyonga P, Warf BC, Kulkarni AV, Monga V, Schiff SJ. Learning based segmentation of CT brain images: application to postoperative hydrocephalic scans. IEEE Trans Biomed Eng. 2018;65(8):18711884.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 9

    Peterson MR, Cherukuri V, Paulson JN, Ssentongo P, Kulkarni AV, Warf BC, et al. Normal childhood brain growth and a universal sex and anthropomorphic relationship to cerebrospinal fluid. J Neurosurg Pediatr. Published online July 9, 2021.doi:10.3171/2021.2.PEDS201006

    • Search Google Scholar
    • Export Citation
  • 10

    Kulkarni AV, Schiff SJ, Mbabazi-Kabachelor E, Mugamba J, Ssenyonga P, Donnelly R, et al. Endoscopic treatment versus shunting for infant hydrocephalus in Uganda. N Engl J Med. 2017;377(25):24562464.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 11

    Schiff S, Kulkarni A, Mbabazi-Kabachelor E, Mugamba J, Ssenyonga P, Donnelly R, et al. Brain growth after surgical treatment of infant post-infectious hydrocephalus in sub-Saharan Africa: two-year results of a randomized trial. J Neurosurg Pediatr. Published online July 9, 2021.doi:10.3171/2021.2.PEDS20949

    • Search Google Scholar
    • Export Citation
  • 12

    Gu C. Smoothing spline ANOVA models. Springer Series in Statistics.Springer;2002.

  • 13

    Pearl J, Mackenzie D. The Book of Why: The New Science of Cause and Effect. 1st ed. Basic Books;2018.

  • 14

    Tingley D, Yamamoto T, Hirose K, Keele L, Imai K. mediation: R package for causal mediation analysis. J Stat Softw. 2014;59(5).doi:10.18637/jss.v059.i05

    • Search Google Scholar
    • Export Citation
  • 15

    Won SY, Konczalla J, Dubinski D, Cattani A, Cuca C, Seifert V, et al. A systematic review of epileptic seizures in adults with subdural haematomas. Seizure. 2017;45:2835.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 16

    Puca A, Fernandez E, Colosimo C, Lauretti L, Pallini R, Tamburrini G. Hydrocephalus and macrocrania: surgical or non-surgical treatment of postshunting subdural hematoma?. Surg Neurol. 1996;45(4):376382.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 17

    Aguiar PH, Shu EB, Freitas AB, Leme RJ, Miura FK, Marino R Jr. Causes and treatment of intracranial haemorrhage complicating shunting for paediatric hydrocephalus. Childs Nerv Syst. 2000;16(4):218221.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 18

    Bin Zahid A, Balser D, Thomas R, Mahan MY, Hubbard ME, Samadani U. Increase in brain atrophy after subdural hematoma to rates greater than associated with dementia. J Neurosurg. 2018;129(6):15791587.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 19

    Miranda LB, Braxton E, Hobbs J, Quigley MR. Chronic subdural hematoma in the elderly: not a benign disease. J Neurosurg. 2011;114(1):7276.

  • 20

    Warf BC. Comparison of 1-year outcomes for the Chhabra and Codman-Hakim Micro Precision shunt systems in Uganda: a prospective study in 195 children. J Neurosurg. 2005;102(4 suppl):358362.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 21

    Bayley N. Bayley Scales of Infant and Toddler Development,. Third Edition. Pearson;2005.

Metrics

All Time Past Year Past 30 Days
Abstract Views 619 476 0
Full Text Views 203 163 20
PDF Downloads 219 156 17
EPUB Downloads 0 0 0