Quantitative detection and follow-up of intracranial hypertension in craniosynostosis: an optical coherence tomography study

Background/purpose To evaluate in craniosynostosis: 1) the diagnostic accuracy of fundoscopy and optical coherence tomography (OCT) to detect intracranial hypertension (ICH); 2) the time course of retinal thickness after treatment of ICH; and 3) the relation between high hyperopia (HH) and fundoscopy/OCT scan findings. Methods Syndromic, multisuture, unicoronal, unilambdoid and sagittal synostosis patients visiting our national center were included in this longitudinal cohort study and formed a consecutive series. Retinal layers on OCT, OCT fundus image and fundoscopy were evaluated. ICH was scored according to presence of abnormal intracranial pressures, hydrocephalus, progressive cerebellar tonsillar herniation or fingerprinting and growth arrest. Diagnostic accuracy of OCT, fundoscopy and fundus image, the time course of retinal thickness after ICH and interference of HH were analyzed using linear mixed models. Results 577 OCT scans in 307 patients were included. ICH was found in 7.2%. Combining total retinal thickness (TRT), OCT fundus image and fundoscopy resulted in a sensitivity of 76% and 81% specificity to detect signs of ICH. TRT was increased in patients who have had signs of ICH versus patients who never had signs of ICH (β+44.9 µm in patients who have had ICH, 95%CI 9.0-80.8,P=0.01). TRT decreased to normal in the years after surgery (β -3.6 µm/year, 95%CI -7.2 - -0.05, P=0.047). There were greater odds of having increased TRT in patients with HH (OR 2.9, 95%CI 1.1-7.6,P=0.03).


Introduction
The detection of intracranial hypertension (ICH) in craniosynostosis is one of the main challenges in managing these patients.Untreated ICH causes developmental problems and may even lead to loss of vision.[3] The best technique for diagnosing ICH in craniosynostosis is 24-hour invasive intracranial pressure (ICP) monitoring. 4However, this investigation needs a surgical procedure, thereby making it unfeasible for routine screening.Alternative methods for detecting ICH include evaluating the presence of risk factors for, and indirect indicators of ICH.6][7] Indirect indicators include symptoms such as morning headaches and behavioral changes, fingerprinting on skull X-ray [8][9][10][11] and papilledema on fundoscopy.
Unfortunately, the assessment of fingerprinting is observer-dependent.There are also problems with the diagnostic accuracy of what may appear to be "papilledema" 4,[9][10][11][12][13][14] .For example, the presence of high hyperopia (HH) may interfere with the detection of papilledema, resulting in a mimic, so-called pseudopapilledema.[17][18][19] Therefore, we aimed to evaluate the following in children with craniosynostosis: 1) the diagnostic accuracy of fundoscopy and OCT in detecting the presence of ICH; 2) the time course of retinal thickness after the treatment of ICH; and 3) whether the presence of HH interfered with fundoscopy/OCT findings of papilledema.

Methods
All patients with syndromic, multisuture, unicoronal, unilambdoid and sagittal synostosis who were managed between 2016-2020, and who were able to comply with OCT scanning, were prospectively included.We also added patients who underwent OCT scanning before 2016 because of a clinical suspicion of ICH.All patients underwent cranial vault expansion in the first year of life, except for patients with a late referral. 6Patients with a late referral only underwent surgery when signs of ICH were present, or in case of severe distortion of the skull shape.The Erasmus MC IRB approved this study (MEC-2015-638; MEC-2017-1143).

OCT protocol
The Spectralis OCT scanner (Heidelberg Engineering, Dossenheim, Germany, Heyex software v6.7.13) was used to obtain OCT scans.For each OCT scan, two parameters were evaluated: 1) TRT; and 2) OCT fundus image.The image of the optic nerve head (ONH) was evaluated by an experienced ophthalmologist (SL).Papilledema was defined as 360 degrees of edema of the optic disc or blurring of the optic disc margins with obscuration of blood vessels.To measure the TRT of the ONH, the ONH was centered, after which the retinal image was focused to optimize scan quality. 19,20An automatic segmentation algorithm included in the Heyex software detected the TRT reference layers (inner limiting membrane and Bruch's membrane).A circular chart with concentric rings at 1, 2 and 3 mm was positioned over the ONH, dividing the ONH in 8 equal quadrants.The mean thickness of the TRT was calculated with a precision of 0.2 µm.OCT scans in which less than 75% of the areas were available were excluded.According to a previous study using the same OCT device on children (age 4-10 years), a TRT of > 503.6 µm was considered as increased, and thus as papilledema. 19Since no TRT normative values exist for the Spectralis device in adults, the adults were scored according to these values as well.

Fundoscopy
Results of all fundoscopies performed during follow-up were extracted from the medical records.
Fundoscopies were scored by experienced ophthalmologists (SL) as either "papilledema" or "no papilledema".Presence of HH or optic disc drusen was considered with the interpretation of the fundoscopy.
Patients underwent fundoscopy regularly according to our follow-up protocol: Irrespective of age, additional fundoscopy was performed when ICH was suspected.

Refractive error
The spheric cycloplegic refractive error (RE) was measured by an experienced orthoptist and was expressed in diopters (D).HH was defined as a spheric equivalent of >+4 D.

Intracranial hypertension
Presence of ICH was scored as being absent or present, which was determined by the following criteria: A) Invasive 24-hour ICP monitoring.The baseline findings were classified as: <10mmHg, normal; 10-15mmHg, borderline abnormal depending on the height and duration of abnormal plateau waves (see below); >15mmHg, abnormal.The trend in ICP values were also checked for any increase overnight.Abnormal ICP plateau waves were categorized as: plateau height <25mmHg (normal), or 25-35mmHg (borderline), or >35mmHg (abnormal); and plateau duration <10 minutes (normal), or 10-20 minutes (borderline) or >20 minutes (abnormal). 21,22 Progressive ventriculomegaly (i.e., hydrocephalus) and obliterated subarachnoid spaces on cerebral magnetic resonance imaging (MRI) scans.C) Progressive cerebellar tonsillar herniation (and/or syrinx) and obliterated subarachnoid spaces on cerebral MRI scans.D) Progressive fingerprinting on skull X-rays or scalloping of the inner cranial cortex on CT scans, and reduced size of the ventricles and/or obliterated subarachnoid spaces.E) Occipitofrontal head circumference (OFC) growth curve deflection, indicating craniocerebral disproportion.6 Presence of ICH was determined in one of three ways: if Criterion A was abnormal; or if Criterion B or Criterion C were abnormal; or if Criterion D and Criterion E were abnormal.
Since the diagnostic accuracy of papilledema was evaluated, papilledema itself was not taken into consideration for the evaluation of the presence of ICH.

Statistical analysis
Statistical analyses were performed using R statistical software (v4.0.3) and consisted of two major parts: 1) diagnostic accuracy (primary outcome), and 2) longitudinal follow-up (secondary outcome).The interference of HH was evaluated as a tertiary outcome.Statistical significance was set at a P-value of < 0.05.

Diagnostic accuracy.
To examine the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio and negative likelihood ratio for fundoscopy findings, OCT fundus image and TRT to detect ICH, we used data collected from the eye with the thickest TRT.
In patients  who developed ICH, the OCT-scan and fundoscopy at the first moment that ICH was detected were used.
 who never had ICH, who also had identical results for each scan or fundoscopy (i.e., all were true negative or false positive), the first scan or fundoscopy was selected for analyses.In patients with contrasting results (i.e., true negative results in one scan or fundoscopy and false positive results in the other), the false positive scans and fundoscopies were selected.
The odds ratios of an abnormal fundoscopy, OCT fundus image or TRT to detect ICH was analyzed by calculating odds ratios from generalized estimation equation (GEE) models, using the cutoff value from normative values and correcting for age as a covariate and multiple 8

A C C E P T E D
measurements (geepack package v1.3-1). 19For this analysis, all OCT-scans were included up until the first moment that ICH was detected.The necessity of interaction terms was tested by evaluation of the quasi-likelihood under independence model criterion (QIC) with and without interaction terms, and interactions were subsequently not entered into the models.

Longitudinal follow-up
The eye with the thickest TRT on the first OCT-scan was selected for all consecutive measurements.Longitudinal time course of retinal thickness after the treatment of ICH was evaluated with 2 linear mixed models (nlme package v3.1-150).Effect plots were generated with the ggplot2 package, v3.3.2.We used three statistical models to examine these data.In the first model, all OCT-scans in all patients who underwent surgery were included.TRT was defined as a continuous outcome variable, whereas a historical finding of ICH (yes/no) and age (continuous) were entered as independent predictors.The necessity of an interaction term between ICH and age was tested by evaluation of the Akaike Information Criterion (AIC) with and without interaction terms, and an interaction was subsequently entered into the first model.
In the second model, all patients who had had papilledema and ICH were included.TRT was defined as a continuous outcome, time after surgery in years was included as a continuous independent variable.In the third model, all patients were included.TRT in patients who did not have ICH but underwent surgery according to our protocol was compared to TRT in patients who did not undergo surgery because of late referral and no signs of ICH.In the statistical model, TRT was defined as a continuous outcome variable, whereas surgery (yes/no but no ICH before surgery) and age (continuous) were the independent predictors.Again, the necessity of an interaction term between surgery and age was tested by evaluation of the AIC with and without interaction terms, and an interaction was subsequently entered into the first model.

Interference of HH
The interference of HH on fundoscopy and OCT results was analyzed by calculating odds ratios from GEE models.For this analysis, all OCT-scans were included up until the first moment that ICH was detected.We used three models.In these models, papilledema on fundoscopy (1) or OCT fundus image (2) and increased TRT (3) were defined as a binary outcome (yes/no), and high HH (yes/no) and ICH (yes/no) were the independent predictors.

Results
The total study cohort included 307 patients (46% female) (Table 1).An attempt to perform an OCT scan failed in 5 patients: 3 had Apert syndrome, one isolated unicoronal synostosis, and one patient had multisuture synostosis.Reasons to not attempt OCT scanning in children included a lack of focus, intellectual disability and a (time-related) lack of parental consent.577 OCT scans were carried out, and the median age at OCT scan was 7.7 years (IQR 5.5-11.2). Figure 1 depicts the in-and exclusion criteria per individual analysis (i.e.diagnostic accuracy as the primary outcome, longitudinal follow-up as the secondary outcome and the interference of HH as a tertiary outcome) and the numbers of included scans and patients per analysis.A total of 283/307 (92.2%) patients underwent surgery, and the median age at first surgery was 0.7 years (IQR 0.5-1.0).Twenty-four patients did not have ICH and did not undergo surgery, 22 because of late referral and 2 because of parental choice.A subset of 131/307 (42.7%) patients underwent repeated OCT scanning, with the median number of 3 scans in these patients (IQR 2-4, range 2-11), and the median follow-up period between the first and last OCT was 2.1 years in this specific subset of patients (IQR 1.0-4.4,range 0.1-9.8).Manual corrections of the segmentation lines for the TRT were performed in 69/577 (12%) OCT-scans.
An increased TRT was diagnosed in 43/237 (18.1%) patients.The OCT fundus image was abnormal in 31/237, and fundoscopy in 29/237.Diagnostic accuracy was evaluated for the TRT, OCT fundus image and fundoscopy separately, for all 3 items combined, and for TRT and OCT fundus image combined (Table 3).In this analysis the pre-test probability of signs of ICH was 7.2% (~1-in-14), and the post-test probability of ICH using all three diagnostic evaluations combined was 23.8% (~1-in-4).

Odds ratios OCT/fundoscopy
360 OCT-scans in 237 patients were analyzed.There was a 13-fold, or 15-fold, greater odds of having ICH when TRT was increased, or papilledema present on fundoscopy, respectively (Table 4).

Course of TRT by age in operated patients
A total of 510 OCT-scans were carried out in 278 patients (Figure 1).56 patients had a history of ICH.TRT was increased in patients who have had ICH versus patients who never had ICH (β+44.9µm in TRT, 95%CI 9.0-80.8).Moreover, the TRT slightly decreased as patients aged (β-4.0 µm/year, 95%CI -5.8 --2.2, Figure 2).The observed decrease by age was not different in patients who have had ICH, versus patients who never had ICH (β+0.4,95%CI -2.7-3.5).

Course of TRT after surgery in patients who have had papilledema and ICH
A total of 119 scans in 40 patients were used including only patients who have had papilledema and ICH (Figure 1).These included 18/119 measurements performed after surgery according to protocol, 77/119 after repeated surgery because of recurrent ICH, and 24/119 because of ICH in patients with a late referral.TRT significantly decreased in the years after surgery (β-3.6 µm/year, 95%CI -7.2--0.05).

Course of TRT in patients who did not undergo surgery versus patients who did undergo surgery but never had ICH
In this analysis, 412 scans were carried out in 250 patients (Figure 1).TRT in the 24 patients who did not undergo surgery was comparable to TRT in patients who underwent surgery according to the protocol, but never had ICH (β +35.2 µm, 95% CI -10.6-81.1),see Figure 3.

Interference of HH on fundoscopy and TRT
Last, we analyzed 307 OCT scans carried out in 199 patients (Figure 1).In total 46/199 (23.1%) had HH.There was close to 3-fold greater odds of having an increased TRT if HH was present, corrected for the true presence of ICH (OR 2.9, 95%CI 1.4-7.6).In contrast, there did not appear to be greater odds of having papilledema on the OCT fundus image or fundoscopy in patients with HH, corrected for the true presence of ICH (OR 2.3, 95%CI 0.8-6.7,and OR 3.1, 95%CI 0.9-9.8).

Discussion
Our study provides three key observations about quantitative detection and follow-up of ICH in patients with craniosynostosis.

Diagnostic accuracy
Following a sensitivity of 76% for the combination of OCT and fundoscopy, approximately 24% of patients do not develop papilledema and/or an increased retinal thickness in the presence of signs of ICH (false negative measurements).
The sensitivity of 59% for fundoscopy alone is remarkably higher compared to the 11-22% reported in literature. 10,16In contrast, the sensitivity on OCT parameters seems to be slightly lower compared to the numbers in the study of Swanson et al. 16 Explanations for this difference include the alternative method to diagnose ICH (i.e., the ICP was measured under general anesthesia and for one minute), and the other OCT parameters which were used to evaluate retinal thickness.The difference between the fundoscopy sensitivity reported by Tuite et al and the current study includes the different patient population (i.e. the majority of their population had non-operated isolated craniosynostosis) and the alternative method to assess ICP (i.e.only baseline ICP was evaluated). 10Besides, the recorded time of the ICP was only 3 hours in some patients, whereas the recommended standard is a 24-hour overnight measurement 4 .

Longitudinal follow-up
This study shows that TRT slightly decreases as children age, and this was comparable in children who have had ICH versus children who never had ICH.Patients who have had ICH kept a significantly increased TRT compared to patients who never had ICH, which is consistent with a previous study from our group. 20After surgery, TRT quickly decreases below the upper limit of normal values. 19We hypothesize that in some patients the retinal thickness remains in the upper bound of normal because of permanent anatomical changes. 23 did not identify a difference in TRT in late referred patients without ICH who never underwent surgery and patients who underwent surgery according to the treatment protocol, but 13

A C C E P T E D
who never had ICH.This finding most likely reflects a mild phenotype in patients with a late referral.It appears that a conservative approach to surgical intervention in this particular group is safe, although long-term assessment of cognition and psychosocial wellbeing should be undertaken.

Influence of HH
HH did not influence the OCT fundus image or fundoscopy outcomes, only TRT was influenced by this refractive anomaly.HH is commonly associated with a crowded optic nerve head, which causes pseudopapilledema. 24The association between an increased TRT and HH reflects both a strength and weakness of TRT, as the precision of the measurement up to a few microns likely results in earlier detection of retinal changes, but it also causes more false positive measurements.These findings represent an additional argument to combine the OCT scan with fundoscopy.Also, total retinal volume, might be of added value to distinguish papilledema from pseudopapilledema in patients with an increased TRT. 25 Caution with the interpretation of this tertiary analysis is warranted, the confidence intervals of the OR were wide.

Clinical diagnosis of ICH
The practice in our center is to check for signs of ICH at every visit, besides fundoscopy and OCT, including complaints of headaches in the morning, changes in behavior or sleep pattern, and decline of OFC.If ICH is suspected, we proceed with imaging (CT and/or MRI) and, sometimes, invasive ICP measurement.The current study underlines the importance of additional diagnostic aids for accurately screening for ICH.Although the pre-to post-test increase in probability of ICH being present was modest (7.2 to 24%, or 1-in-14 to 1-in-4) for certainty in identifying ICHthey are better than what we have at the moment using clinical 14 A C C E P T E D assessment alone.26 Our study does have limitations.First, only 17 patients were diagnosed with ICH, which resulted in pre-test probability for signs of ICH of only 7.2% and limited the evaluation of sensitivity.This low number may be related to our protocol with routine surgery within the first year of life.Also, we only had invasive ICP measurements in 8 patients, and the majority of the cases with ICH were diagnosed based on proxy measurements.Invasive measurements represent the best technique for diagnosing ICH in craniosynostosis, but are unfeasible for routine screening since it requires a surgical procedure and inpatient care for 24 hours.The calculated sensitivity and specificity should therefore be interpreted with caution.However, we did use several other clinical derivatives of and risk factors for ICH to evaluate the presence of ICH, such as ventriculomegaly, Chiari I malformation and skull growth arrest.These individual risk factors potentially suffer from measurement error, but pooling them together reduces the chance of missing or over-diagnosing ICH.Also, we analyzed the diagnostic accuracy based on a group of patients with isolated and syndromic craniosynostosis who are at risk of developing ICH.This may have influenced our calculated parameters, since the risk of ICH varies within these diagnoses.For future studies, we would recommend an analysis per diagnosis, in patients with a low and high risk of developing ICH, within a greater sample size.Also, we did not have enough data to analyze the diagnostic accuracy and longitudinal follow-up of the peripapillary RNFL, an OCT parameter which has been studied by others. 16,18[29] 15

A C C E P T E D Conclusion
The sensitivity of TRT, OCT fundus image and fundoscopy to detect the presence of signs of ICH is 71%, 59% and 59% respectively and 76% if the three methods are combined.We found that the presence of HH affects the OCT findings and needs to be considered when evaluating OCT results.An increased TRT (after excluding HH) and papilledema on fundoscopy, combined with a clinical suspicion of ICH, warrants further screening for the presence of ICH.During follow-up after surgery, the increased TRT due to ICH remains higher compared to the TRT of age-matched patients without ICH, but falls within the bounds of normative values.
Sagittal synostosis: preoperatively, and yearly from the age of 2-6 years  Unicoronal and unilambdoid synostosis: preoperatively, and at the age of 2, 3 and 4 years  Crouzon syndrome: preoperatively, every 3 months from the age of 1-2 years, biannual from the age of 2-5 years, and annual from the age of 5-10 years  Apert and Saethre-Chotzen syndrome and multisuture craniosynostosis involving the lambdoid suture(s): preoperatively, biannual from the age of 1-5 years and annual from the age of 5-10 years  Muenke syndrome, complex craniosynostosis without lambdoid synostosis, IL11RA and HUWE1 gene mutation craniosynostosis: preoperatively, biannual from the age of 1-2 years, and annual from the age of 2-10 years.

Figure 2 .
Figure 2. Effect plot of linear mixed model, visualizing the course of total retinal thickness in

Figure 3 .
Figure 3.Effect plot of linear mixed model, visualizing the course of total retinal thickness in Fig 1

Table 1
Patient characteristics.

Table 2
Classification of intracranial hypertension

Table 3
Diagnostic accuracy of total retinal thickness, OCT fundus image and fundoscopy to detect intracranial hypertension.

Table 4
Odds of having intracranial hypertension in patients with an increased TRT or papilledema on OCT fundus image or fundoscopy (dichotomous outcomes).
2Odds of having ICH if one of the following parameters indicated ICH: 1) TRT; 2) OCT fundus image; 3) fundoscopy2Odds of having ICH if one of the following parameters indicated ICH: 1) TRT; 2) OCT fundus image Abbreviations -ICH: intracranial hypertension, SE: standard error, TRT: total retinal thickness