CROP – The Clinico-Radiologico-Ophthalmological Paradox in Multiple Sclerosis: Are Patterns of Retinal and MRI Changes Heterogeneous and Thus Not Predictable?

Background To date, no direct scientific evidence has been found linking tissue changes in multiple sclerosis (MS) patients, such as demyelination, axonal destruction or gliosis, with either steady progression and/or stepwise accumulation of focal CNS lesions. Tissue changes such as reduction of the retinal nerve fiber layer (RNFL) and the total macular volume (TMV), or brain- and spinal cord atrophy indicates an irreversible stage of tissue destruction. Whether these changes are found in all MS patients, and if there is a correlation with clinical disease state, remains controversial. The objective of our study was to determine, whether there was any correlation between the RNFL or TMV of patients with MS, and: (1) the lesion load along the visual pathways, (2) the ratios and absolute concentrations of metabolites in the normal-appearing white matter (NAWM), (3) standard brain atrophy indices, (4) disease activity or (5) disease duration. Methods 28 MS patients (RRMS, n = 23; secondary progressive MS (SPMS), n = 5) with moderately-high disease activity or long disease course were included in the study. We utilised: (1) magnetic resonance imaging (MRI) and (2) -spectroscopy (MRS), both operating at 3 Tesla, and (3) high-resolution spectral domain-OCT with locked reference images and eye tracking mode) to undertake the study. Results There was no consistency in the pattern of CNS metabolites, brain atrophy indices and the RNFL/TMV between individuals, which ranged from normal to markedly-reduced levels. Furthermore, there was no strict correlation between CNS metabolites, lesions along the visual pathways, atrophy indices, RNFL, TMV, disease duration or disability. Conclusions Based on the findings of this study, we recommend that the concept of ‘clinico-radiologico paradox’ in multiple sclerosis be extended to CROP–‘clinico-radiologico-ophthalmological paradox’. Furthermore, OCT data of MS patients should be interpreted with caution.


Introduction
A current and frequent topic of ophthalmological debate is whether the reduction of the retinal nerve fiber layer (RNFL) or the total macular volume (TMV) has any consistent correlation with brain-or spinal cord atrophy, or even the brain weight of multiple sclerosis (MS) patients. If a definitive correlation does exist in the majority of MS patients, the underlying pathogenic mechanisms remain unclear, and would require elucidation. It is difficult to imagine how the 1-3 million axons of the optic nerve (which is itself part of the central nervous system (CNS)) should reflect billions of axons within the CNS. Such neuronal convergence would suggest pathological processes other than solely 'trans-synaptical neurodegeneration'.
A correlation between brain atrophy and any other clinical or physiological parameter in MS patients has yet to be identified (e.g. [1][2][3]). Despite ongoing investigations, the cause of MS is unknown and each patient follows his/her individual course. Axonal injury or axonal dysfunction is generally accepted to be the pathological correlate to temporary dysfunction or disability in MS patients. But neither high field magnetic resonance imaging (MRI), magnetic resonance spectroscopy (MRS) or high-resolution spectral domain-optical coherence tomography (SD-OCT) (which allows visualisation of axonal changes in the CNS or the retina in vivo), enable a reliable prediction that could be applied to individual MS patients. On the contrary, more than two decades of MS research has led to the concept of 'the clinico-radiological paradox' and 'radiological isolated syndrome', as neither lesion load, brain volume/atrophy nor any other radiological parameter was found to correlate strictly to the patients' clinical parameters [1][2][3].
A key study combining (1) stringent clinical inclusion criteria, (2) advanced high-resolution OCT, and (3) MRI/MRS technique, in order to investigate whether a 'clinico-radiological-ophthalmological paradox' (CROP) exists in MS has yet to be performed, and was the aim of our study.

Participants
This study was approved by the local Ethics Committee (Commission of Medical Ethics of Vienna-Ethics Approval/Registration Numbers: EK 06-169-VK and EK-08-028-0308; and Ethical commission of the Medical University of Vienna-Ethics Approval/Registration Number: 414/2008). Informed, written consent was obtained from all patients and volunteers before entry into the study.
All MS patients included in this study (and in two previously published studies [33][34]) had to fulfil Poser's diagnostic criteria [35], Barkhof's MRI criteria [36] and had to have detectable oligoclonal bands [37]. 59 MS patients consented to the use of high-resolution spectral domain-OCT (SD-OCT) [33], 37 MS patients consented to MRI and magnetic resonance spectroscopy (MRS), both operating at 3 Tesla [34], but only 28 MS patients consented to both OCT and MRI/MRS. 23 patients followed a relapsing-remitting course with well-defined relapses and lack of clinical progression between relapses (RRMS, mean age, 38 Table 1).
All MS patients had their MRI/MRS and OCT examination on the same day, one scan immediately after the other. Prior to investigations, hydration of patients (a confounding factor) was ensured by the intake of sufficient amounts of water (approximately 1 to 2 litres). Detailed patient demographic and clinical data has been described previously [38][39] and again in Table 1. Patients were treated with beta-interferons, glatiramer-acetate, mitoxantrone and natalizumab (Table 1).

MRI and MRS
The standardized MRI/MRS protocols were carried out as described previously [34].

Analysis of Lesion load along the visual pathways
Lesion regions were analyzed with T 2 -TIRM axial multi slice MR images with a spatial resolution of 0.45 x 0.45 x 4 mm, using the medical imaging software package Jim (Xinapse Systems Ltd, Aldwincle, UK), by two independent assessors (IK, LV) with experience in MRI image analysis. Image slices were inspected between the first appearance of the inferior horns of , drug withdrawal 20 months before OCT examination; 5 , high titres of anti-interferon autoantibodies, drug withdrawal 14 months before OCT examination; 6 , mitoxantrone cumulative dose 96mg per m 2 body surface, drug withdrawal 10 months before OCT examination; lateral ventricles, and the point where the posterior horns of lateral ventricles lost their sharpness. Lesions were semi-automatically selected along the visual pathways. Assessors manually located lesions based on the difference in signal intensity between lesions and surrounding tissue, with the software automatically outlining the entire lesion. In the case where one lesion was also covering the area outside the visual pathway, those parts of the lesion were manually excluded. For regional quantification purposes the lesions were divided into four groups according to their location in the brain (anterior dextra, anterior sinistra, posterior dextra and posterior sinistra). The posterior edge of the inferior horn of lateral ventricle was set as the border between posterior and anterior regions. The volume of the brain in the same slice range was also selected semi-automatically. The cavities and ventricles were excluded from the whole brain volume. The total lesion load (defined as the lesion volume (ml) over brain volume (ml)) in all evaluated slices, and regional lesion load (lesion load per quadrant) were subsequently calculated. For each parameter, the mean was calculated from two independent examinations performed by two assessors (IJK, VK) who were blind to the data of each other, and without knowledge of the clinical and OCT data.

Analysis of Brain Atrophy
An experienced examiner (HR), without knowledge of the clinical and OCT data, determined the following established indices to quantify global and focal, cortical and subcortical atrophy [40][41][42]: 1) the Evans ratio; 2) the caudate head index (CHI); 3) the basal cistern index (BCI); 4) the cella media index (CMI); 5) the maximum width of the third ventricle; 6) the maximum width of the fourth ventricle; 7) the maximum width of the anterior interhemispheric fissure (MIF); 8) the maximum width of the Sylvian fissure (MSF); and 9) the maximum frontal subarachnoid space (MFSS).

High Resolution spectral domain OCT
We utilised a high resolution SD-OCT which combines OCT technology with a confocal scanning laser ophthalmoscope (Heidelberg Engineering, Heidelberg, Germany, Spectralis software version 4.0.3.0, Eye Explorer Software 1.6.1.0). A special eye-tracking mode (TrueTrack™) and high-scanning speed allows the reduction of artefacts due to eye movement. Each peripapillary OCT is registered and locked to a reference image. OCT software can identify previous scan locations and "guide" the OCT laser beam to scan the identical location again. To optimize the signal-to-noise ratio and image quality, 16 frames (B scans) of the same scanning position were averaged with the Automatic Real-Time averaging mode (ART mode). All RNFL scans were acquired in the high-resolution acquisition mode allowing a more detailed differentiation of retinal layers, with pupil dilation. Furthermore, signal strength has been shown to affect RNFL thickness measurements using conventional Stratus OCT [33; 38]. Therefore, scans with low quality (signal strength for RNFL and TMV scans < 15) and failing RNFL segmentation were excluded. Measurements were repeated until a technically-excellent quality was achieved. Criteria for determining scan quality included: (1) a clear fundus image before and during image acquisition, (2) absence of scan or algorithm failures, (3) even and dense grey scale saturation throughout all retinal layers with dense grey visible in the RPE, and (4) the RNFL visible without missing or blank areas and a continuous scan pattern.
All automated measurements of macular thickness and volume were performed through dilated pupils with a high-resolution macular scan protocol allowing for a more detailed differentiation of retinal layers (the TMV compounds of inner limiting membrane, nerve fiber layer, ganglion cell layer, inner plexiform layer, inner nuclear layer, outer plexiform layer, outer nuclear layer, external limiting membrane, photoreceptor layer, and retinal pigment epithelium).
The SD-OCT imaging protocol comprised 49 B-scans per volume scan of 20°× 20°, and each scan was averaged with 9 frames per B-scan. Topographic macular surface maps were constructed automatically by the OCT software and displayed with numeric averages of the mean thickness for each of the nine map sectors (F, foveal; TI, inner temporal; TO, outer temporal; II, inner inferior; IO, outer inferior; NI, inner nasal; NO, outer nasal; SI, inner superior; SO, outer superior) within three concentric regions of 1, 3, and 6 mm diameter, respectively, as defined by the Early Treatment Diabetic Retinopathy Study (ETDRS).
All RNFL and macular scans were performed several times by one skilled and trained observer (NS) within one session until at least 3 high-quality scans were achieved and used for further analysis. The observer had no knowledge of clinical data or specific baseline and MRI/ MRS data. Final analysis was only performed on scans without segmentation errors, and no manual correction was performed in any case. Normal ranges of RNFL thickness and TMV was determined by the internal database of the OCT device and compared to published reports and prior experience [4; 26].
Patients with other ocular pathologies that may reduce RNFL thickness such as glaucoma, anterior ischemic optic neuropathy, high myopia, and congenital abnormalities of the optic nerves etc., were excluded.

Statistics
This was a prospective study. We used descriptive and analytic statistics as previously described in detail [33; 38]. Briefly, nonparametric tests (Mann-Whitney, Kolmogorov-Smirnov, Bonferroni-Holm, Chi-squared) and regression analysis were applied (statgraphics plus 5.1). All test results were considered significant if p-values were below 0.05. All parameters are expressed as means, medians, minimums and maximums and standard error of the mean. As case numbers are rather low, the data has been presented in detailed surveys (Tables 1-5) and in an individual manner to overcome the limitations of group comparisons (see cases 1-5 in the results and references [33; 38-39]).

Results
Our results may be summarized briefly as follows: We found inter-individual differences in clinical characteristics, NAA levels of the NAWM, the lesion load in the visual pathways, brain atrophy indices and both RNFL thickness and the TMV (data not shown). The levels ranged from normally to markedly reduced levels.
Neither the RNFL nor TMV was found to correlate strictly with: (1) brain metabolites in the NAWM (NAA, Cho and Cr) or (2) the various brain atrophy indices and lesion load in the visual pathways. Evan's ratio, CHI, CMI, BCI were found to be within normal range in all 28 patients even though global brain atrophy was detectable on MRI. This may suggest that these indices were not applicable to the brain atrophy seen in the MS patients in this study ( Table 3). The atrophy detected in MS patients seemed to be described better by other parameters such as the maximum width of 3 rd and 4 th ventricles, MIF, MSF and MFSS (see results and Table 1, patient 1, 24 and 25, Fig 1). Usually the whole brain volume and white and grey matter volume/fractions are determined in MRT studies of MS patients to detect even mild brain atrophy. Also in most of our included MS patients there could at least a mild brain atrophy be detected by inspection of his/her MRI by a skilled radiologist and or neurologist. Interestingly, also in those MS patients with marked brain atrophy also the RNFL or TMV were not significantly reduced (this applies also for the further 2 years follow-up [38; 39]). However, we state clearly that the brain atrophy indices-we have used here-have potential methodological limitations and basically do not correlate strictly with disease course or clinical progression in all MS patients as this has been shown for so many other MRI parameters before [1].
Most importantly, there was no correlation between individual specific sectors of the retina (6 retinal sectors and globally (Table 5)) and the lesion load of their corresponding visual pathways. Correlations and possible associations were analysed for total groups of RRMS and SPMS patients (Table 6), but checked intra-individually for all 28 MS patients for plausibility and stringency. This means that the values for each patient were intentionally broken down and traced back, i.e. each individual RNFL value (global or the corresponding sector) has been analysed set to the corresponding visual pathway lesion load, the brain metabolite levels and brain atrophy indices. Nevertheless, we generally did not find consistent patterns that could suggest retro-or anterograde trans-synaptical degeneration, either in patients with, or more importantly, in patients without previous ON. (It should however be noted that our study design did not allow the measurement of whole brain volume and the parenchymal fractions for the white and grey matter, and tractography of the posterior visual pathway.) On the other hand, each MS patient seemed to follow his own, very individual pattern. In order to make this clear, we present very briefly five patients (2 RRMS without ON, 1 RRMS with ON, 1 SPMS patient with ON and 1 SPMS without ON).  Firstly, a female RRMS patient without previous ON and moderately long disease duration (6 years), but (1) very active disease course, (2) highest visual pathway-lesion load in the RRMS group without ON, (3) lowest ratio of NAA to Cr (1.46) and second-lowest absolute concentration of NAA in her NAWM (7.08mM), and (4) signs of brain atrophy with highest values for the maximum width of her 3 rd and 4 th ventricles (9.5 and 14.0mm), MFSS (3.0), MIF (4.0) and MSF (5.0), showed normal values for RNFL and TMV. In other words, one would expect very low RNFL and TMV values in this patient, not the opposite, i.e. normal values, which are partly the highest of all included MS patients (RNFL global, right eye, 110 μm; left eye, 101μm). The TMV-values were 8.0mm³ (right eye) and 8.36 mm³ (left eye). This patient had an EDSS of 3.0. Before entry into our study, this patient was treated with mitoxantrone and interferons, and further along the disease course, natalizumab therapy had to be established due to ongoing high disease activity (patient 1, Table 1; and Fig 2).
Secondly, a similar female RRMS patient (  1, patient 1). It should be noted that patient 7 is 7.5 years older than patient 1, but had nearly the same lesion load in her visual pathway as patient 1, with (1) an absolute concentration of NAA in her NAWM (11.97 mM) and a ratio of NAA to Cr (1.98) within normal range, and no other signs of brain atrophy. Patient 7 had an EDSS of 4.0. This patient also had to be treated with natalizumab later in the disease course due to high disease activity (Table 1). Thirdly, a female RRMS patient with previous ON of her right eye (patient No 20, Table 1), a rather long and active disease course (5.0 years), but otherwise with normal findings for absolute concentration of NAA in NAWM (13.85 mM), a ratio of NAA to Cr of 2.07 and no signs of brain atrophy with normal brain atrophy indices and the lowest visual pathway lesion load of all 28 included MS patients, had a significantly reduced RNFL on her right eye compared to her left eye (global, 87 μm vs. 100 μm; T, 67 μm vs. 86 μm; TS, 148 μm vs. 165 μm; TI, 110 μm vs. 159 μm; I, 100 μm vs. 125 μm). Interestingly, the TMV was within normal range and showed no side difference (right eye, 8.44 mm³ vs. left eye, 8.49 mm³). Why the RNFL reduction in the temporal sectors of the patient's right eye was not associated with the TMV in this patient remains unclear. Her EDSS was 1.5.
Fourthly, a male SPMS patient (patient 26, Table 1) with a very long disease duration (25 years) and long periods of highly active disease course with prior ON of each eye had unexpectedly normal values for RNFL (global: right eye, 111 μm; left eye 103 μm) and TMV (right eye, 8.1 mm³; left eye, 7.95 mm³). This seems even more paradoxical as: (1) brain atrophy was obvious, (2) the brain atrophy indices were partly out of range (maximum width of 3 rd and 4 th ventricles, 8.9 mm and 12.9 mm respectively; MFSS 3.0; MIF 2.0; MSF 2.8), (3) NAA levels in the NAWM were markedly reduced (absolute concentration, 8.25 μm), and as (4) the lesion load of the visual pathway was nearly the highest of all included patients (2 nd rank). Of note is that the RNFL on both sides was within 'normal range', but significantly reduced only in his left eye compared to the fellow eye, although he experienced ON in both eyes. In any case, the RNFL and TMV were found to be reduced compared to the patient's right eye. (Reliable normative databases do not exist for RNFL or TMV, and the individual baseline levels before disease onset is of course unknown). His EDSS was 6.5. Finally, and in clear contrast to the last patient described, the fifth (male SPMS) patient (patient 24, Table 1) did not experience previous ON but was of similar age as patient 4 at entry to the study (46.5 years). While the total disease duration was significantly less (6.5 years vs. 25 years), disease activity and disability was comparable high (EDSS 6.5). Brain atrophy was less obvious, the brain atrophy indices were only partly out of range, but less so compared to the last patient (maximum width of 3 rd and 4 th ventricles, 8.7 mm and 10.6 mm; MSF 2.0). The lesion load in the visual pathways was a third less, while the brain metabolites were within normal range (NAA, absolute concentration, 10.59 mM; ratio of NAA to Cr, 1.84). However, RNFL values were significantly reduced (RNFL global, 81 μm and 78 μm). The TMV correlated well to RNFL values (right eye, 8.01 mm³ and left eye, 7.33mm³).
In summary, we found no strict correlation between the specific parameters examined (Fig  1, Table 6). The RNFL thickness of patients' eyes with ON compared to their unaffected fellow eyes, to healthy controls and the internal databases of OCT devices has been found reduced in most, but not all patients [33]. The set of various parameters was complex and heterogeneous, and did not strictly follow a unique distinct pattern, which would have served for further forecast analysis (and prediction).

Discussion
It has long been recognised that neither lesion load nor brain-or spinal cord atrophy correlates strictly to the degree of disability in MS. MS patients with high relapse rates and high lesion load, but only mild impairment are characteristic of the well-known 'clinico-radiological paradox' [3; 43]. The cause and mechanisms for this remain to be elucidated. Furthermore, the  Group  G  G  T  TS  TI  S  I  N  NS  NI  T  TS  TI  S  I  N  NS  NI  OD  OS  OD  OD  OD  OD  OD  OS  OS  OS  OS  OS  OS  OS  OS  OD  OD    This replacement of MRI with OCT is remarkable as only few studies with consistent and comparable study design and methodology have yet been published. Most of the studies (search terms, "OCT", "MS" and "brain atrophy", see above, introduction [6-8; 16; 20-22; 25]) were performed, (1) with conventional time domain OCT [6-8; 16; 21-22], and (2) the case numbers of the studies were generally low, even when they were published in high-ranked journals (ref. 6 [25], 68 RRMS and 9 SPMS).
According to Popper's 'falsifiability' theorem, studies with low case numbers allow the confirmation or rejection of a general hypothesis, and this is suggested to be valid in the case of MS patients. In other words, accurately-derived negative results need not be compensated by a high number of cases: 'one negative result may discard a theory until the opposite is proven', particularly, as the situation in healthy controls of different sex, background and age is not clear. Large normative databases for RNFL and TMV for the different OCT devices are still not available. More importantly, daily physiological variations of the retina might significantly exceed the anticipated annual change of RNFL thickness (0.1-2 μm) [32]. Balk et al. identified hydration status as a crucial confounder for any retinal scan, statistical analysis and further interpretation, even in young healthy controls. Retinal changes were found to be significantly reduced between baseline and first follow-up scans performed before and after a 10km charityrun (dehydration status). To rehydrate, the participants were advised to drink water and sports drinks. Significant RNFL and TMV reduction were found to be fully reversible in the follow-up scans only 1 to 1.5 hours after the 10km charity run. It is generally accepted to refer to this phenomenon as 'RNFL reduction' rather than 'RNFL thinning' [32]. Of note, Balk et al relativized this hydration-/rehydration-effect to <1% of the OCT data variation in a recently published  study [44] when they reinvestigated the 26 healthy test persons of their original study [32]. Balk et al interpreted their new data as follows,"(1) Normal variation of OCT data may mask small degrees of neurodegeneration, (2) Hydration related cellular volume changes may be a cause for OCT data variation, (3) their prospective trial demonstrates that hydration causes < 1% of OCT data variation and (4) Trials using OCT will need to consider normal variation." [44]. Balk et al opened very interesting, new questions such as variations of the macular RNFL, ganglion cell layer, inner plexiform layer, inner nuclear layer, outer plexiform layer and outer nuclear layer, and variations caused by instrument/software related factors [44], which may explain the RNFL variations found in their first study apart from a hydration/rehydration-effect only [32]. It is essential that the study protocol in their second study was changed: they excluded a dehydration of participants, hence only the effects of hydration on the retina were studied, and may explain the partly controversial results of their both studies [32; 44]. To the best of our knowledge, it has yet not been studied in detail and in a large cohort of MS patients if hydration related cellular volume changes may be a cause for OCT data (and/or MRT?) variation. However, it is well known that many MS patients limit their fluid intake to reduce their bladder urge during their daily routine [45].
If current investigations (OCT, MRI or MRS) are still too imprecise to detect and monitor the subtle, early changes that exist in MS patients (or that cannot be separated from gross ON changes), it should at least be possible to detect the hypothesized changes in all remaining MS patients with high activity, long disease courses and obvious progression (progressive relapsing MS, SPMS and PPMS). According to this hypothesis which is referred to by several studies, many patients would indeed escape detection in very early disease stages or if the disease follows a benign course [6-8; 16; 20-22; 25]. The OCT data published to date indicate to a new paradox: 'CROP-the clinico-radiologico-ophthalmological paradox'.
The aim of our detailed study was to confirm or discard CROP. The power of our study was that: (1) MRI/MRS and OCT were performed on the same day immediately after each other, (2) that the MS patients were encouraged to drink sufficiently before the examinations [32], (3) that advanced high-resolution SD-OCT technique and (4) MRI/MRS technique was used on (5) well classified MS patients with partly high disease activity and/or long disease courses, and finally, (6) that all data are traceable to allow readers their own interpretation [36-37; 39]. Most importantly, the disease activity remained high in further follow-up appointments with a median observation period of 22 ± 0.5 months, but showed neither significant reduction of the RNFL nor the TMV [36-37; 39]. This is of greater interest as we balanced our 'inclusion criteria' to 'false negative' rather than 'false positive' inclusion criteria; in other words, we included only patients that would be classified as 'MS patients' no matter which diagnostic criteria would have been used.
However, we could not identify a strict correlation or rule that applied to all MS patients which allows a reliable prediction either of RNFL/TMV reduction over time, other examined parameters or disease course. This might be explained by: (1) MS patients that follow a benign course, (2) relapsing-remitting courses without any disease progression (and pathological changes) between relapses [3-4; 43], (3) focal lesions in the spinal cord or certain regions in the brain such as the brain stem or cerebellum etc. which correlate better to physical impairment and hence, the EDSS, (4) methodological limitations that make it impossible to detect all ongoing, subtle changes under the detection limits (less than ±2-4μm RNFL) of available SD-OCT devices [4-5; 25-27], (5) subtle changes that cannot be separated from gross pathological changes by focal lesions in the retina or anterior visual pathway [20; 25-27], or [6] that other, as of yet undefined pathogenetic mechanisms of degeneration and repair affect the CNS tissue differently (e.g. [45][46][47][48]).
We conclude that the well-known 'clinico-radiological paradox' or 'radiological isolated syndrome' [3-4; 43] also applies to the OCT, i.e. that every patient seems to follow his own individual course with an individual set of para-clinical parameters. As others we think that it is to premature to suggest OCT as surrogate marker or tool to measure and monitor cerebral and spinal atrophy [49][50].