A M E R I C A N J O U R N A L O F O T O L A R Y N G O L O G Y – H E A D A N D N E C K M E D I C I N E A N D S U R G E R Y 3 5 ( 2 0 1 4 ) 1 9 2 – 1 9 7

Ava i l ab l e on l i ne a t www.sc i enced i r ec t . com

ScienceDirectwww.e l sev i e r . com/ loca te /amjo to

Is a redefinition of the growth criteria of vestibularschwannomas needed?

Dan Hougaard, MDa,⁎, Anne Norgaard, MDb, Thue Pedersen, MDb,Bo Martin Bibby, MSc, PhDc, Therese Ovesen, DMSci a

a ENT Department, Aarhus University Hospital, Denmarkb Department of Neuroradiology, Aarhus University Hospital, Denmarkc Department of Biostatistics, Aarhus University, Denmark

A R T I C L E I N F O

⁎ Corresponding author. ENT Department, AaTel.: +45 61689062.

E-mail address: ddh@dadlnet.dk (D. Houg

0196-0709/$ – see front matter © 2014 Elsevihttp://dx.doi.org/10.1016/j.amjoto.2013.08.002

A B S T R A C T

Article history:Received 22 May 2013

Purpose: The natural history of vestibular schwannomas is poorly understood. Knowledgeof growth rate and growth pattern is essential because the treatment strategy is based uponthese. The purpose of this study was to determine the inter- and intraobserver variability inmeasuring VS size.Materials and methods: Two consultant neuroradiologists independently made three linearmeasurements (d1, d2, d3) using digital MRI scans. MRI scans from 72 patients diagnosedbetween 2002 and 2010 with VS were obtained. These patients had a total of 223 MRI scans.d1 (medio-lateral diameter) was made perpendicular to d2. d2 was made parallel to theposterior border of the petrous ridge, and d3 was a measure of the cranio-caudal height ofthe tumor.Results: Limits of Agreement ranges are larger for interobserver reliability compared tointraobserver reliability. Measurement error for all diameters (except d1, intraobserver) isgreater than 2 mm. d1 measurements had the least variability and d3 measurements thehighest variability, both for intra and interobserver measurements.Conclusions: The optimal method of estimating VS size needs further investigation, andmeasurements need to be standardized and clearly defined. d3 seems to be the mostdifficult diameter to measure reliably. Interobserver measurement error for all diameters isgreater than 2 mm. The current VS growth criterion of more than 1–2 mm, used to triagepatients to surgery, lies within this error range, and thus is problematic as a guide for clinicalpractice. We therefore suggest that the growth criterion for VS be redefined.

© 2014 Elsevier Inc. All rights reserved.

1. Background

Unilateral vestibular schwannoma (VS) is a benign tumorarising from abnormally proliferative Schwann cells, whichenvelope the lateral portion of the vestibular nerve in theinternal acoustic meatus [1]. The etiology of VS is poorlyunderstood. Genetic defects, such as chromosomal 22q de-

rhus University Hospital

aard).

er Inc. All rights reserved

letions [1], and infectious causes, such as herpes virus, havebeen implicated in the pathogenesis of VS [2].

With the exception of patients with type II neurofibroma-tosis, no genetic predisposition for VS has been identified.Neurofibromatosis patients, in contrast to those with sporadicSchwannomas, usually develop bilateral VS.

VS typically presents with asymmetrical, high-tone sen-sorineural hearing loss (75% of cases) together with a

, Building 10, 1st Floor, Norrebrogade 44, 8000 Aarhus, Denmark.

.

193A M E R I C A N J O U R N A L O F O T O L A R Y N G O L O G Y – H E A D A N D N E C K M E D I C I N E A N D S U R G E R Y 3 5 ( 2 0 1 4 ) 1 9 2 – 1 9 7

discriminatory loss. Sudden deafness is the initial symptomin about 10% of cases and the remaining 15% present withdisequilibrium, unilateral tinnitus, cranial nerve symptomsor Ménière's disease-like symptoms [3].

The growth pattern of VS is multifaceted. VS may growcontinuously or only to a certain size, followed by stagnationor even shrinkage [4]. Growth of VS leads to gradual filling ofthe internal acoustic meatus and, in some cases, protrusionout of the porus. Growth occurs within the initial 5 yearsfollowing diagnosis; growth occurs in a limited number oftumors and primarily in those with an extra-meatal extension[4]. VS are for the most part slow growing tumors, but so far ithas not been possible to predict the individual growth pattern.[1,5,6] Many studies have attempted to identify predictors oftumor growth, but have failed to identify any correlationbetween growth and variables such as age, laterality, sex andduration of symptoms [6]. One study found unilateral tinnitusto be predictive of tumor growth [7], while another has foundtumor size at diagnosis to be predictive of tumor growth [8].

Knowledge of the VS growth rate and growth pattern isessential because the treatment strategy is based upon this [4].In Denmark, VS surgery is centralized and performed at onlyone facility; Copenhagen University Hospital. Surgical in-dications include tumor size larger than 15 mm, tumor growthof more than 2 mm, and cystic tumors which may displaysudden and dramatic growth [4]. If these criteria are not met,the patients are allocated to a “wait and scan” regimen inwhich the audiovestibular symptoms and VS growth aremonitored regularly. VS growth is assessed by MRI (Fig. 1).

Reporting the size of VS is a challenge, which is reflectedby the multitude of different classifications currently in use(at least 20 classifications are available) [1]. Gadolinium-enhanced MRI is the gold standard imaging investigation for

Fig. 1 – VS treatment st

tumors of the cerebellopontine angle, with a sensitivity ofalmost 100% and an extremely high specificity, capable ofdetecting tumors as small as 4 mm [9]. The traditional methodof assessing tumor size is a single (axial) linear measurementassessing the greatest diameter. This method is also the onecurrently used in Denmark to register growth of VS [4,10,11].However, several studies suggest that a more accurate way ofdescribing VS growth is by volumetric measurement [12,13]. VStumors grow in all directions and may be shaped in differentways, ranging from spherical to lollipop-shaped to dumbbell-formed. The shapemay also change as the tumor expands intothe intracranial space [13,14].

The incidence of VS in Denmark has been increasingduring the last four decades. The annual number of diagnosedVS has increased from 15 cases in 1976 to 123 cases in 2004,corresponding to an increase in incidence from 3.1 VS permillion per year to 22.8 VS per million per year in 2004. After2005 a slight decrease in the number of diagnosed tumors hasbeen noted, with 105 cases in 2008 (corresponding to anincidence of 19.4 VS per million per year) [10].

1.1. Hypothesis

A single (axial) linear measurement may not be the best way ofassessing VS tumor size and growth. Growth occurs in threedimensions and the shape of the individual tumor is unpre-dictable [13–15]. Furthermore, other studies suggest that thisway of measuring VS is subject to significant intra- andinterobserver variability [15,16]. One study has shown aninterobserver variability of 7.4 mm (Limits of Agreement, LoA)and an intraobserver variability of 2.5 mmwhenmeasuring thesame tumor [16]. Yet,manyauthors consider a 1-2 mmincreasein tumor size on a subsequent scan, based on two-dimensional

rategy in Denmark.

194 A M E R I C A N J O U R N A L O F O T O L A R Y N G O L O G Y – H E A D A N D N E C K M E D I C I N E A N D S U R G E R Y 3 5 ( 2 0 1 4 ) 1 9 2 – 1 9 7

measurements, as evidence of growth [4,14,16]. We thereforehave reason to believe, that there is significant variability inmeasuring the size and evaluating the growth of VS.

1.2. Objective

To determine intra- and interobserver variability for differentmeasurements of VS size, at the Department of Neuroradiol-ogy, Aarhus University Hospital, Denmark.

1.3. Study design

Retrospective analysis of all patients with MRI verified VS.Primary outcome was the intra- and interobserver variationwhen measuring VS.

1.4. Population

All patients seen between 2002 and 2010, at the ENTDepartment, Aarhus University Hospital, Denmark, with adiagnosis of Vestibular Schwannoma were included. Patientsrecords were identified by searching for the ICD code DD33.3B,corresponding to a vestibular schwannoma diagnosis. Thispopulation of patients includes both “wait and scan” patients,as well as patients who had VS surgery. From 2002 andonwards digitalMRI images were used on a regular basis at theDepartment of Neuroradiology, Aarhus University Hospital.

A total of 128 patients were identified. 56 of these patientswere excluded, predominantly due to having only intra-meatal tumors. Others were excluded because digital imagesof the MRI scans were not available, the diagnosis waswithdrawn after revision of the images, the images werepurely done postoperatively, and in one case because thesame patient appeared twice. We also excluded patients withVS containing cystic components (Fig. 2).

For detailed baseline characteristics of the patients includ-ed refer to Table 1.

Fig. 2 – Trial profile.

1.5. Measurements

Measurements were done on T1 weighted images with gado-linium contrast (Dotarem 0.5 mmol Gd/ml, Guerbet Group).

All examinations at the Department of Neuroradiology,Aarhus University Hospital were performed using either a 3Tesla Achieva Philips MRI or a 1.5 Tesla MAGNETOM AvantoSiemens MRI. A small number of scans were performed atother locations using different MRI scans. Seven patients hadCT scans instead of MRI, but these were excluded.

Slice intervals were, using the 3 Tesla MRI, 1 mm recon-structed as 0.5 mm slices covering a 38 mm area over pori in acranio-caudal direction, and 0.7 mm slices covering a cranio-caudal area of 50.4 mm using the 1.5 Tesla MRI.

Transversal T2 FLAIR sequence was performed at theinitial MRI scan. Transversal T2 weighted sequence was alsoperformed initially and carried out at repetitive MRI scans.When planning sequences both the coronal- and sagittalplanes were included. The slices were made perpendicular tothe internal ear and outer ear canal bilaterally, using thecoronal plane and parallel to the lower edge of CorpusCallosum when looking at the sagittal plane. When contrastwas given, a transversal T1 weighted sequence was per-formed. All images were interpreted by two experiencedconsultant neuroradiologists independently (observer 1 andobserver 2). Both neuroradiologists re-measured 16 randomlyselected MRI scans from 16 patients, in order to evaluate theintraobserver variability.

Both observers made linear measurements as follows:

a) Medio-lateral diameter (axial slice, d1): from the medialedge of the tumor to an imaginary line drawn along theposterior surface of the temporal bone diameter,parallel to the petro mastoid part of the temporal bone(to exclude the intra meatal component).

b) Antero-posterior diameter (axial slice, d2): parallel tothe petrous ridge.

c) Cranio-caudal height (coronal slice, d3).

Observer 1 and 2 measured all VS scans in these threedimensions. Both were blinded to each other's measurementsand to earlier descriptions and former measurements of theMRI scans.

Table 1 – Population characteristics.

Characteristics

PatientsTotal 72Age at 1st MRI, yMean (SD) 57.6 (11.33)Sex, number (%)Male 41 (56.9)Female 31 (43.1)MRI scansTotal 223Average per patient (range) 3.1 (1–8)SurgeryNumber of patients (%) 13 (18.1)

Table 2 – Estimates and tests for systematic differencesbetween observer 1 and observer 2.

Estimate P value

d1, (mm, ± 95% CI) 0.33 ± 0.20 .0008d2, (mm, ± 95% CI) 0.43 ± 0.28 .002d3, (mm, ± 95% CI) 0.87 ± 0.22 <.0001Volume, (mm3, ± 95% CI) 97 ± 58 .001

195A M E R I C A N J O U R N A L O F O T O L A R Y N G O L O G Y – H E A D A N D N E C K M E D I C I N E A N D S U R G E R Y 3 5 ( 2 0 1 4 ) 1 9 2 – 1 9 7

After thorough examination of d1, d2 and d3 we made theassumption, thatVShasanellipsoidal shapewhenexcluding theintrameatal portion (see Figs. 3A and 3B). Therefore the volumewas calculated using the formula: V = (pi/6) × d1 × d2 × d3.

1.6. Statistics

The diameters d1, d2, and d3 as well as the volume wereanalyzed using Bland-Altman plots and paired t-tests to test forsystematic variation within or between the two neuroradiolo-gists. More specifically, this was done using a linear mixedmodel with radiologist as a fixed effect and the patient, theinteraction between patient and radiologist, and the scan asrandom effects. The data were analyzed using R version 2.15.0.

2. Results

We compared measurements of VS size performed by twoneuroradiologists on MRI scans. We found that there weresignificant systematic differences between observer 1 andobserver 2 for all three diametersmeasured (Table 2). Observer2 had a tendency to measure all diameters larger thanobserver 1. However, the most significant systematic differ-ence between the two observers was for d3 (Fig. 4).

Bland Altman plots were used to visualize the interobserverdifferences. 95 % of the measured differences were within theLoA. Table 3 shows, as expected, that all LoA ranges are widerwhen considering interobserver vs. intraobserver variability. All

Fig. 3 – A & B: Measuring of vestibular schwannomas. Figure3A visualizes d1 (B) and d2 (A). Figure 3B visualizes d3 (A).

three diameters had LOA greater than 2 mm for interobservervariability. d1 had the lowest LOA and d3 the highest LOA.

3. Discussion

ENT doctors frequently face a dilemma when managing VSpatients: to perform surgery or to allocate patients to a waitand scan regime? Symptomatic patients with large or cystictumors need active treatment. However, it is less clearwhether patients with small or medium sized tumors, whichare asymptomatic, require active treatment. Tumor growth,if any, is unpredictable. One third of VS are purely intrameatal at the time of diagnosis [4] and intra meatal tumorsdo not impose a life threatening risk to the patient. Becausesymptoms, rather than size, initiate surgical treatment andbecause a consensus meeting in Japan in 2003 definedgrowth of purely intra meatal tumors as growth to an extrameatal extent, we excluded these tumors. For two-dimen-sional VS size measurements many authors state that a 1–2 mm increase in subsequent scans is evidence of growth[4,12,16]. Our study shows, that this value lies within thelimits of agreement, for both intra and interobservervariability (except d1 LoA, intraobserver). As expected, wefind greater agreement in intraobserver measurementscompared to interobserver measurements. In our study wefind d1 to have the smallest LoA, both intra and inter-observer. However, one study has shown interobserverdifferences as high as 7.46 mm when measuring themedio-lateral diameter (d1) [17]. In Denmark an increase ofmore than 2 mm when measuring VS using the longestsegment in the axial plane defines growth. It is not clearlydefined if this is merely a difference seen between twoconsecutive scans or if this is calculated as an annualgrowth rate [4]. Our study emphasizes the importance ofstandardizing and clearly defining the measurements relat-ed to VS growth/size. When looking at all three diametersobserver 2 consistently measured higher than observer 1;especially for d3. One reason for this could be that theheight of the tumor is more difficult to assess. Although thesystematic difference was small, it was significant, becauseof the large number of MRI scans included. We did notconsider having a third neuroradiologist do the measuringbecause both our radiologists were equally experienced andalso received their training at the same facility. In agree-ment with our findings, Marshall et al. concluded thatgrowth of 2 mm cannot be reliably measured [17], and thesefindings are consistent with what we find. Another studyshows that the maximum diameter along the pyramid is abetter parameter than calculating the volume [12]. A

Fig. 4 – Bland-Altman plots. Interobserver variability. x-axis: average measurement, y-axis: measurement differences, solidblack lines: 95% reference intervals, dotted black line: mean difference.

196 A M E R I C A N J O U R N A L O F O T O L A R Y N G O L O G Y – H E A D A N D N E C K M E D I C I N E A N D S U R G E R Y 3 5 ( 2 0 1 4 ) 1 9 2 – 1 9 7

different study, using area-tracing software to determine thevolume, concludes that volume measurements are moreaccurate compared to two-dimensional measurements forthe evaluation of VS growth [14]. Yet another study suggeststhat the maximum diameter method, commonly used todetermine growth, is the least reliable measurement method[13]. This is mainly due to a high degree of intraobservervariability and systematic underestimation of the volume ofsmall tumors and overestimation of the volume of largetumors. This study recommends that multiple dimensionsbe measured [13]. Currently there is no golden standard formeasuring VS growth, but according to The AmericanAcademy of Otolaryngology Head & Neck Surgery guidelines,VS growth should be measured on the best axial imageusing two linear measurements (parallel and perpendicularto the petrous ridge) [18].

Table 3 – Limits of Agreement, lowest intraobservervariability and highest interobserver variability shownin bold.

Intraobserver Interobserver

d1, mm ± 1.93 ± 2.29d2, mm ± 2.57 ± 2.77d3, mm ± 2.25 ± 3.09Volume, mm3 ± 373 ± 1203

In order to facilitate and standardize the measuring of VS,one could argue that specific MRI protocols should beintroduced. Thickness of cuts and standardized positioningof patients in regards of temporal bone planes will impact theappearance of tumors when comparing serial imaging. Astandardization of protocols throughout the entire countrywith regards to type of MRI machine, thickness of cuts, andplanes in relation to the temporal bone would of course beaspirational, but in reality very hard to introduce.

When allocating patients to a wait and scan regime, onemust bear inmind that thepatients are committed to aprogramof surveillance, using MRI scans, for an undetermined period oftime. During this period of time the patients are stigmatizedwith an intra-cranial tumor which may exhibit growth.

A priori, the experience of the MRI technician and theradiologist, the type of MRImachine, the use of contrast agent,the ability of the patient to lie still, and the image acquisitionprotocol all play important roles in the quality of the image.Studies have shown, however, that there is negligiblevariability between different types of MRI machines [18].

Our study involved a comparison of three different linearmeasurements to assess VS size, performed by two experi-enced neuroradiologists on 223 MRI scans. A possible weak-ness of the study is the assumption of VS being ellipsoidal, asno two VS have exactly the same shape. Furthermore,volumetric measurements are hard to compare betweendifferent studies. When using volumetric measurements the

197A M E R I C A N J O U R N A L O F O T O L A R Y N G O L O G Y – H E A D A N D N E C K M E D I C I N E A N D S U R G E R Y 3 5 ( 2 0 1 4 ) 1 9 2 – 1 9 7

variability of each individual diameter is multiplied, thusincreasing themeasurement error.And asnoted the calculationof a volume also requires that a “standard” shape for VS isdefined. Given the variability in the three linear measure-ments (and hence in the volume) for both inter andintraobserver measurements, our study does not find volu-metric measurements to be superior to linear measurements.In theory, the volume (size and shape) would provide moreaccurate dimensions of VS. Stereological methods, whereseveral cuts are used to reconstruct the 3-D structure of anobject, could be a means of analyzing the entire VS moreaccurately. This method, however, would also be associatedwith both intra- and interobserver variability, as this methodwould require both subjective measuring and subjective shapetracing to some extent.

Measurements made by different observers are lessreproducible that those made by the same observer [15,17].To test the true intraobserver variation in this study, the twoneuroradiologists involved would have had to re-measure all223 MRI scans in three different planes. Therefore we insteadre-measured 16 randomly selected VS patients in all threeplanes. These findingswere consistent with other studies thatfound LoA to be smaller when looking at intraobserverdifferences (vs. inter) [15,17]. One must bear in mind, though,that part of this can be explained by the systematic differencein measuring between the two neuroradiologists. Whenlooking at the current clinical practice, a lot of differentradiologists do the measuring of VS at many differentlocations. We therefore mainly focused on the inter observervariation, as we believe, this way of measuring has the mostaccurate correlation to the daily clinical practice.

The LoAs found in this study (all but one) exceed the currentVS growth criterion of 1–2 mm for allocating patients to surgeryvs. “wait and scan” approach. To provide evidence of tumorgrowth the measurement ought to at least exceed the limits ofagreement. Accordingly we suggest that the current growthcriterion needs to be redefined. We further recommend, basedon our findings, that VS growth surveillance be done at largercentralized facilities by a limited number of highly experiencedneuroradiologists and that further investigation is needed todevelop better methods for VS growth/size determination.

In order to truly evaluate our way of measuring and theclinical decisions that would have been imposed as aconsequence hereof, one could argue that a comparison ofthe following should be done: comparison of the categoriza-tion (growth/no growth) and clinical decisions (surgery/nosurgery) imposed as a consequence of either measuring thelargest diameter in one plane (currently done at one facility inDenmark) and the measuring of either all three diameters orfor instance d1 (which in our study had the lowest LoA).

4. Conclusion

Intraobserver variability in measurements of VS tumor size issmaller than interobserver variability. Importantly, all interob-server Limits of Agreement exceed 2 mm. d1 (medio-lateraldiameter) has the smallest variability in regards to both inter-and intraobserver variability, and d3 seems to be the mostdifficult diameter to measure reliably. Accordingly, we recom-

mend thatVSgrowth surveillancebeperformedat large facilitiesby a limited number of highly trained neuroradiologists, that thebest way of measuring VS needs further investigation, thatmeasurementsought tobestandardizedandclearlydefined, andthat the current growth criterion of 1–2 mmormore needs to beredefined to exceed the LoA. We suggest that VS growth shouldinstead be defined as a 3 mm linear increase in d1 on twoconsecutive MRI scans one year apart.

R E F E R E N C E S

[1] Tos M, Natural history of Schwannoma. Scott-Brown’sOtorhinolaryngology, Head and Neck Surgery 7th edition2008:3957–66. [BOOK].

[2] Burkhart CG. Herpes and acoustic neuromas: is there a causeand effect to observe? Med Hypotheses 2010;74:1013–4.

[3] Gimsing S. Vestibular schwannoma: when to look for it?J Laryngol Otol 2010;124:258–64.

[4] Stangerup SE, Caye-Thomasen P, Tos M, et al. The naturalhistory of Vestibular Schwannoma. Otol Neurotol 2006;27:547–52.

[5] Beenstock M. Predicting the stability and growth of acousticneuromas. Otol Neurotol 2002;23:542–9.

[6] Nikolopoulos TP, Fortnum H, O'Donoghue G, et al. Acousticneuroma growth: a systematic review of the evidence. OtolNeurotol 2010;31:478–85.

[7] Agrawal Y, Clark JH, Limb CJ, et al. Predictors of vestibularschwannoma growth and clinical implications. Otol Neurotol2010;31:807–12.

[8] Artz JC, Timmer FC, Mulder JJ, et al. Predictors of futuregrowth of sporadic vestibular schwannomas obtained byhistory and radiologic assessment of the tumor. Eur ArchOtorhinolaryngol 2009;266:641–6.

[9] Newton JR, Shakeel M, Flatman S, et al. Magnetic resonanceimaging screening in acoustic neuroma. Am J Otolaryngol2010;31:217–20.

[10] Stangerup SE, Tos M, Thomsen J, et al. True incidence ofvestibular schwannoma?Neurosurgery 2010;67:1335–40[discussion 1340].

[11] Stangerup SE, Caye-Thomasen P, Tos M, et al. Change inhearing during 'wait and scan' management of patientswith vestibular schwannoma. J Laryngol Otol 2008;122:673–81.

[12] Fiirgaard B, Pedersen CB, Lundorf E. The size of acousticneuromas: CT and MRI. Neuroradiology 1997;39:599–601.

[13] Varughese JK, Wentzel-Larsen T, Vassbotn F, et al. Analysis ofvestibular schwannoma size in multiple dimensions: acomparative cohort study of different measurementtechniques. Clin Otolaryngol 2010;35:97–103.

[14] Van de Langenberg R, de Bondt BJ, Nelemans PJ, et al. Follow-upassessmentof vestibular schwannomas: volumequantificationversus two-dimensional measurements. Neuroradiology2009;51:517–24 [Epub 2009 May 6].

[15] Cross JJ, Baguley DM, Antoun NM, et al. Reproducibility ofvolume measurements of vestibular schwannomas – apreliminary study. Clin Otolaryngol 2006;31:123–9.

[16] Fucci MJ, Buchman CA, Brackmann DE, et al. Acoustic tumorgrowth: implications for treatment choices. Am J Otol 1999;20:495–9.

[17] Marshall AH, Owen VM, Nikolopoulos TP, et al. Acousticschwannomas: awareness of radiologic error will reduceunnecessary treatment. Otol Neurotol 2005;26:512–5.

[18] Slattery 3rdWH, Fisher LM, Yoon G, et al. Magnetic resonanceimaging scanner reliability formeasuring changes in vestibularschwannoma size. Otol Neurotol 2003;24(4):666–70.