Clinical correlates of longitudinal MRI changes in CADASIL

Abstract

Previous studies showed that various types of cerebral lesions, as assessed on MRI, largely contribute to the clinical severity of CADASIL. However, the clinical impact of longitudinal changes of classical markers of small vessel disease on conventional MRI has been only poorly investigated. One hundred sixty NOTCH3 mutation carriers (mean age ± SD, 49.8 ± 10.9 years) were followed over three years. Validated methods were used to determine the percent brain volume change (PBVC), number of incident lacunes, change of volume of white matter hyperintensities and change of number of cerebral microbleeds. Multivariable logistic regression analyses were performed to assess the independent association between changes of these MRI markers and incident clinical events. Mixed-effect multiple linear regression analyses were used to assess their association with changes of clinical scales. Over a mean period of 3.1 ± 0.2 years, incident lacunes are found independently associated with incident stroke and change of Trail Making Test Part B. PBVC is independently associated with all incident events and clinical scale changes except the modified Rankin Scale at three years. Our results suggest that, on conventional MRI, PBVC and the number of incident lacunes are the most sensitive and independent correlates of clinical worsening over three years in CADASIL.

Inroduction

Cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL) caused by mutations of the NOTCH3 gene is the most frequent hereditary cerebral small vessel disease (SVD).¹ The main manifestations develop over decades and include attacks of migraine with aura, mood disturbances, ischemic strokes, and progressive cognitive decline.²

There is accumulating evidence indicating that various cerebral markers of SVD on conventional magnetic resonance imaging (MRI) delineated according the STRIVE criteria³ largely contribute to clinical worsening in CADASIL. White matter hyperintensities (WMH) are mainly related to speed cognitive processing from the early stage of the disease.^4,5 The number of lacunes predict the increase of disability and occurrence of dementia over three years.⁶ Incident lacunes are also associated with a rapid decline in executive performances.⁷ Cerebral microbleeds (MBs) are related to poor functional outcome at cross-sectional level⁸ and seem to predict few aspects of clinical degradation at longitudinal level.⁶ Finally, brain atrophy, that seems to result from the accumulation of lacunes at subcortical level,^9,10 was also found strongly related to disability and global cognitive performances in a follow-up study of 76 patients.¹¹

However, the exact clinical meaning of longitudinal MRI changes has been only poorly investigated in CADASIL. This information is obviously crucial for selecting the best imaging markers as potential outcomes in future clinical trials. Herein, we aimed to determine how the changes of these SVD MRI markers are associated with clinical worsening over a three-year period in CADASIL.

Materials and methods

Participants

Patients were prospectively recruited between September 2003 and April 2011 through a prospective CADASIL study obtained in two centers in France and Germany.⁶ In all subjects, the diagnosis of CADASIL was confirmed by genetic testing. All patients underwent both MRI and clinical examination at study inclusion and three-year follow-up. Details of the study protocol have been reported elsewhere.⁸

Clinical and demographic data were collected at study entry and included age, sex, years of education, and vascular risk factors. Patients with diagnosis of hypertension, diabetes or hypercholesterolemia were treated according to the current recommendations. A history of transient ischemic attacks (TIAs) or stroke and dementia was obtained at baseline and follow-up. Subjects underwent a detailed neurological and neuropsychological examination at baseline and follow-up. Global cognitive function was assessed using the Mattis Dementia Rating Scale (MDRS). The initiation/perseveration subscale of the MDRS (MDRS-I/P) was analyzed separately. Executive functions were assessed by the time to complete Part B of Trail Making Test (TMTB). Part A of the same test (TMTA) was used in parallel to evaluate simple speed processing. Disability was assessed using the modified Rankin Scale (mRS). Moderate or severe disability was defined as mRS score ≥3. Functional independence was determined using the Barthel index. Written informed consent was obtained from all participants or a close relative if the patient was too severely disabled. This study was approved by the ethics committees of both participating institutions (Paris: Ile de France X, No. AOR02-001 and Munich: the LMU medical faculty, No. 299/03). This study was conducted in accordance with the principles of the Declaration of Helsinki of 1975 (revised 1983) and with the guidelines for Good Clinical Practice.

MRI acquisition and image processing

Scanner details and MRI sequences have been reported elsewhere.⁹ The protocol included millimetric three-dimensional (3D) T1-weighted images, axial slices of 5 mm thickness of FLAIR images, and T2*-weighted gradient-echo images. Brain MRI was performed at baseline and during follow-up using the same protocol in each center. Analysis of MRI markers was done blindly to the clinical history.

WMH were analyzed on FLAIR images. All FLAIR axial slices from the base of the cerebellum to the vertex were analyzed. Masks of lesions were generated from FLAIR images by applying a threshold on signal intensity and corrected manually. Then, the total volume of WMH was recorded for each exam. The total volume of WMH was normalized to the intracranial cavity (ICC) in each patient [normalized WMH volume = (WMH volume/volume ICC)×100] (percentage of ICC).

Baseline lacunes were segmented manually using 3D T1 images by two trained raters. Hypointense lesions on T1-weighted images, with a signal identical to cerebrospinal fluid, sharp delineation and a diameter >2 mm, were selected for segmentation as previously reported.⁷ Special care was taken to exclude enlarged perivascular spaces. Incident lacunes were identified using difference imaging from 3D T1 images obtained at baseline and follow-up as already detailed.¹⁰

MB was defined as rounded foci ≤5 mm in diameter hypointense on gradient-echo sequences distinct from vascular flow voids, leptomeningeal hemosiderosis, or non-hemorrhagic subcortical mineralization. The location and number of MB were recorded separately for each exam.

Determination of the global brain volume at baseline was performed as previously detailed.⁹ The brain parenchymal fraction (BPF) was defined as the ratio of brain tissue volume to the ICC volume. Changes of the global brain volume between two time points (percent brain volume change, PBVC) were calculated and checked visually for all subjects using SIENA,¹² an algorithm already used in CADASIL.¹¹ PBVC values are negative and decrease when brain atrophy increases.

Statistical analysis

Incident strokes were assessed irrespective of whether the subject already experienced a stroke. Incident dementia and incident moderate or severe disability were considered only in individuals without dementia or without moderate or severe disability (mRS 0-2) at baseline. Variations of incident clinical events at follow-up were tested with McNemar tests. Differences in clinical scales and MRI parameters between baseline and follow-up were tested with paired samples. Wilcoxon tests were used when these variables were non-normally distributed.

Multivariable logistic regression analyses were performed to assess the independent association between the changes of different MRI markers and the incident clinical events. To assess the relative and independent contribution of changes of different MRI markers on the changes of different clinical scales during follow-up (defined by the difference between 3y and baseline scores), we used mixed-effect multiple linear regression analyses. Some variables were log-transformed before entering the model to meet the normal distribution of residuals. Multiple linear regression analyses were further adjusted for the following fixed effects: age and sex, and for center as a random effect.

Tests were two-sided. The significance level was fixed at 5%. All statistical analyses were performed with R software (https://www.r-project.org).

Results

Cohort demographics

Among the 190 patients who had a complete set of MRI sequences (both at baseline and three years) and high-quality images, 30 individuals who had their MRI with a different scanner at the second time point in the German center were excluded to avoid inter-scanner variability (particularly in the quantitative measures of brain atrophy¹³), leaving 160 patients for analysis. The baseline characteristics of this population are detailed in Table 1.

Table 1.

Main clinical and MRI features at baseline.

	Baseline	Three-year FU
	n = 160	n = 160
Characteristics
Age, mean (SD)	49.6 (10.7)
Female, n (%)	84 (52.5)
Level of education,  median [IQR]	4 [4–6]
Cardiovascular risk factors, n (%)
Hypertension	27 (18.2)
Hypercholesterolemia	60 (40.8)
Active smoking	30 (20.4)
Alcohol consumption	13 (8.1)
Diabetes mellitus	3 (2.0)
Main clinical manifestations, n (%)
Asymptomatic	11 (6.9)
Migraine with aura	85 (53.1)
TIA	46 (28.8)
Stroke	75 (46.9)	96 (60.0)
Gait disturbance	33 (20.6)
Balance problems	38 (23.8)
mRS ≥3	18 (11.2)	25 (15.6)
Dementia	15 (9.4)	32 (20.0)
Clinical scales, median [IQR]
mRS	0 [0–1]	1 [0–2]
MDRS	141 [137–144]	142 [134–143]
MDRS-I/P	37 [34–37]	37 [32–37]
Barthel index	100 [100–100]	100 [100–100]
TMTA time	39 [29–55]	36 [26–55]
TMTB time	92 [62–137]	83 [67–137]
MRI markers, median [IQR]
No. of lacunes	2 [0–6]	5 [0–8]
nWMH	5.49 [3.15–9.94]	6.53 [3.76–11.14]
No. of MB	0 [0–1]	0 [0–2]
BPF	87.0 [83.1–90.0]	86.1 [80.9–89.2]

FU: follow-up; TIA: transient ischemic attack; mRS: modified Rankin scale; MDRS: Mattis Dementia Rating Scale; MDRS-I/P: initiation/perseveration subscale of the MDRS; TMTA: Part A of Trail Making Test; TMTB: Part B of Trail Making Test; nWMH: normalized volume of white-matter hyperintensities; MB: microbleeds and BPF: brain parenchymal fraction.

Clinical changes and variations of MRI markers over three years

The mean time (±SD) between the baseline and the latest follow-up examination was 3.1 ± 0.2 years. After three years, incident stroke (all were presumably of ischemic type in the absence of recent hemorrhage on CT-scan or MRI) occurred in 21 of 160 patients (13.1%). Moderate or severe disability occurred in 7 of 142 subjects (4.9%) and incident dementia occurred in 17 of 145 patients (11.7%). At the end of follow-up, a significant deterioration was detected for most clinical scales when compared with baseline measurements (mRS: +0.33 ± 0.84, P < 0.001; MDRS score: −2.25 ±8.0, P < 0.05; MDRS-I/P: −1.3 ± 3.5, P < 0.001; Barthel index: −2.5 ± 10.5, P < 0.001).

At three-year follow-up, a significant increase of the number of lacunes, volume of WMH and number of MB (number of incident lacunes: 0.4 ± 0.7, P < 0.001; increase of normalized volume of WMH: 0.88% of ICC, P < 0.001; increase of number of MB: 1.5 ± 5.0, P < 0.001) was observed. A significant brain atrophy was also measured over the same period (PBVC: −1.4% ± 1.5, P < 0.001). The three-year follow-up characteristics of the clinical scales and MRI markers are detailed in Table 1.

Independent association between variations of MRI markers and clinical changes over three years

The links between changes of different MRI markers and clinical manifestations over the three-year follow-up period are shown in Table 2. Incident strokes were found independently associated with incident lacunes (OR = 2.1; 95% CI: 1.1-3.8) and brain atrophy (OR = 0.7; 0.5-0.99). Incident dementia and occurrence of moderate or severe disability were significantly associated with brain atrophy (OR = 0.5; 0.4-0.8 and OR = 0.5; 0.3-0.8, respectively). After adjustment for age, sex and center effect, the results did not change (data not shown).

Table 2.

Multivariate logistic regression analyses of incident events (bold significant results at p<0.05).

	Incident strokea		Incident dementiab		mRS ≥ 3c
	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P
No. of incident lacunes	2.1 (1.1–3.8)	0.016	1.6 (0.8–3.2)	0.16	1.6 (0.6–4.0)	0.36
Δ nWMH	1.0 (0.6–1.6)	0.96	1.3 (0.7–2.2)	0.39	0.7 (0.2–2.1)	0.51
Presence of incident MB	0.3 (0.1–1.1)	0.064	0.6 (0.1–2.3)	0.44	0.3 (0.03–3.3)	0.34
PBVC	0.7 (0.5–0.99)	0.048	0.5 (0.4–0.8)	0.0007	0.5 (0.3–0.8)	0.0048

OR: odds ratio; CI: confidence interval; mRS: modified Rankin scale; nWMH: normalized volume of white-matter hyperintensities; MB: microbleeds and PBVC: percent brain volume change.

^aIn all 160 patients.

^bIn 145 patients without dementia at baseline.

^cIn 142 patients with a Rankin scale <3 at baseline.

The association between the different MRI markers and clinical scales variations over three years is shown in Table 3. After adjustment for age, sex and center effect, the decrease of MDRS, MDRS-I/P and Barthel index and increase of TMTA were found independently associated with brain atrophy (standardized ß: 0.37, P < 0.001; 0.25, P = 0.004; 0.21, P = 0.02 and −0.26, P = 0.003, respectively). Increase of TMTB was independently associated with brain atrophy and incident lacunes (standardized ß: − 0.30, P = 0.001 and 0.26, P = 0.004, respectively). The mRS was independently associated with increase of WMH lesions (standardized ß: −0.21, P = 0.01).

Table 3.

Association between changes of MRI markers at three-year follow-up and changes of clinical scores at three years.

	Δ mRS		Δ MDRS		Δ MDRS-I/P		Δ Barthel Index		Δ TMTA		Δ TMTB
	Std ß	P	Std ß	P	Std ß	P	Std ß	P	Std ß	P	Std ß	P
No. of incident lacunes	0.10	0.26	0.09	0.29	−0,06	0.48	−0.14	0.10	−0.03	0.76	0.26	0.004
Δ nWMH	−0.21	0.01	−0.01	0.94	0.04	0.64	0.08	0.34	0.10	0.24	−0.01	0.89
Δ No. of MB	−0.03	0.75	0.01	0.89	0.04	0.66	−0.03	0.76	−0.06	0.52	−0.02	0.86
PBVC	−0.06	0.50	0.37	<0.001	0.25	0.004	0.21	0.02	−0.26	0.003	−0.30	0.001
r2	0.081		0.224		0.208		0.080		0.139		0.196

mRS: modified Rankin scale; MDRS: Mattis Dementia Rating Scale; MDRS-I/P: initiation/perseveration subscale of the MDRS; TMTA: Part A of Trail Making Test; TMTB: Part B of Trail Making Test; Std ß, standardized ß; nWMH: normalized volume of white-matter hyperintensities; MB: microbleeds and PBVC: percent brain volume change.

Note: Adjusted for age and sex as fixed effect and center as a random effect.

Discussion

The major findings of this study can be summarized as follows. First, within a cohort of 160 CADASIL patients, a significant progression of brain atrophy and increase of subcortical cerebrovascular lesions can be easily measured over three years of follow-up. This is observed parallel to 13% incident strokes, 12% incident dementia and 5% incident cases with moderate or severe disability. Second, among the different imaging markers assessed on conventional MRI, brain atrophy and incident lacunes are the two markers significantly related to the clinical changes observed at end of follow-up.

The changes of the number of lacunes, MB, volume of WMH or of the whole brain volume assessed in this large sample of CADASIL patients are important. The annual incidence of lacunes of about 9% is more than twice that observed in elderly population-based studies.^14,15 The mean annualized rate of normalized WMH volume increase is 0.28% of ICC (mean absolute WMH volume increase equals 4016 mm³), 10 times higher than that measured in subjects with manifest arterial disease from the SMART-MR study.¹⁶ The incidence of MB is 10.5% per year, three times that detected in healthy elderly subjects in the Rotterdam study,¹⁷ while the mean annualized rate of brain volume loss was 0.45%, more than twice that usually measured in ageing populations.¹⁸ These variations appear significant from 18 months of follow-up based on data obtained in a subgroup of 144 patients who had two MRI examinations and prior to most clinical scores that change only at three-year follow-up (additional data in Supplementary Table 1). The lower sensitivity of clinical scores to the progression of the disease in comparison to MRI markers might have different sources. Important changes of performances may occur at individual level according to the time elapsed since the last focal cerebral insult. Learning and ceiling effects with repeated testing can reduce the ability to detect small cognitive changes. Finally, the clinical impact of incident cerebrovascular lesions may also become significant only after the accumulation of a minimal level of lesions. Hence, simple imaging quantitative markers obtained on conventional MRI should really be considered as potential surrogates to evaluate early the progression of the disease in future clinical trials.

Among the different MRI markers, the number of incident lacunes is found independently associated with incident stroke and increase of TMTB time during follow-up. This is in line with a previous longitudinal study focusing on incident lacunes in a different sample of individuals from the same cohort.⁷ However, brain atrophy is found associated with much more clinical outcomes at three years. Incident stroke, incident dementia and incident cases with moderate or severe disability are significantly related to the progression of brain atrophy. Moreover, most clinical scales measuring global cognition, executive performances and functional independence vary significantly with the reduction of brain volume. These results are in line with studies in CADASIL showing that among all MRI markers, brain atrophy is one of the most strongly correlated with clinical severity.^6,19 They confirm data obtained in a previous longitudinal study over two years showing that brain atrophy, when measured alone, correlates with the progression of disability and global cognitive decline.¹¹ They are also in accordance with the results of Liem et al.²⁰ who found that the enlargement of ventricles was associated with decreased executive performances after seven years in a small group of CADASIL patients. In the present study, the highest number of clinical correlates with brain volume changes may be explained not only by the high robustness and reproducibility of cerebral volume measurements using the SIENA method¹³ but also because brain atrophy most likely results from multiple key tissue changes. There is accumulating evidence showing that the development of microstructural loss and lacunes at subcortical level has major consequences on the brain morphology and promotes the development of cerebral shrinking in CADASIL^9,10 including cortical atrophy with neuronal apoptosis.²¹ The recent findings of microinfarcts or layers thinning observed at cortical level may further aggravate this phenomenon.^22,23 Conversely, WMH do not necessarily correspond to a reduction of structural tissue components and may even, in some cerebral regions, translate into an increased brain volume.²⁴ Also, incident MB that do not correspond to cerebral tissue loss, may not have direct clinical consequences. Altogether, these findings strongly support that global brain atrophy measurements may be considered as one of the best surrogate endpoint for future therapeutic trials.

In this study, multivariate analysis shows that the increase of WMH is associated with a significant decrease of mRS at three years. These unexpected results should be interpreted cautiously for different reasons: (1) a ceiling effect in the progression of WMH may occur in the most severe patients with extensive WMH at baseline, (2) brain atrophy may counterbalance the extension of WMH and even reduce their volume in some individuals with the fastest clinical progression, (3) the progression of WMH in some areas may remain clinically silent because they do not correspond to myelin or axonal loss and therefore appear protective by contrast to other lesions.²⁵ A further analysis obtained at individual level showed that 10 patients presented a decrease of WMH between baseline and follow-up but that none of them had a decrease of their initial mRS score with half of them having an increase of this score at three years (data not shown). This further emphasizes that increase of WMH volumes should not be considered as a primary imaging marker when the aim is to predict clinical progression over a short time frame. The results might differ in individuals observed at the very early stage of the disease before the occurrence of lacunes and atrophy.

The methodological strengths of this study include the prospective design, large number of patients followed in two different referral centers and in different countries as usually obtained in multicenter clinical trials, and quantitative assessment of WMH volumes and brain atrophy. It is worth noting that SIENA, which corrects for differences in imaging geometry using the outer skull surface for both time points, reduced the effects of scanner drift and inter-scanner variability on longitudinal morphometric results by comparison to using cross-sectional methods at each time point.²⁶ In this study, we even removed patients for which images have not been acquired with the same scanner to further improve the reproducibility.¹³ A potential concern about brain volume measurements might be an effect of time-of-day on brain volume changes as found in elderly subjects with or without dementia.²⁷ We cannot exclude this possibility, but the mean PBVC measured at three years in our CADASIL patients was −1.4%, well above the mean reversible effect of about −0.2% per 12 h in that study. Besides, based on our large population and longitudinal design, it is unlikely that the scan times might not be randomly distributed.

Our study also has limitations. MRI data were not obtained in healthy individuals for better dissecting the normal aging effect on imaging measures. A selection bias is also highly probable because of the exclusion of the most severe cases who could not return to follow-up visits or who were excluded because of head movements and insufficient quality images. Besides, SIENA is well suited to compute the brain volume change based on the movement of image edges between two-time points. It does not consider incident lacunes. Nevertheless, one can imagine that further including incident lacunes in the computation of brain atrophy would increase the actual brain atrophy and thus reinforce our results. Moreover, in view of the major MRI changes observed during CADASIL, these results cannot be transposed to other hereditary small artery diseases or to sporadic SVDs that have very different progression and patterns of lesions. Last, given the small number of incident events, we had to choose a parsimonious model which limited the adjustment and statistical power of our analysis.

In conclusion, this study shows that among the most usual MRI markers of SVD, brain atrophy and in a much lesser extent, the number of incident lacunes appears as the best and most sensitive and independent imaging correlates of clinical worsening in CADASIL. These results may have implications for the choice of candidate surrogate endpoint and/or outcomes in the preparation of future randomized clinical trials in CADASIL.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by grants from the French Ministry of Health (Regional and National PHRC AOR 02-001) and Research (ANR, RHU TRT_cSVD), Association de Recherche en NEurologie Vasculaire, the Vascular Dementia Research Foundation, the Fondation Leducq (Transatlantic Network of Excellence on the Pathogenesis of Small Vessel Disease of the Brain; http://www.fondationleducq.org). Dr Ling is funded by the French Chinese Foundation for Science and Applications (FFCSA) and the China Scholarship Council (CSC).

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Authors’ contributions

YL, FDG, EJ, MDu, MDi and HC designed the study. FDG, EJ, MDu, DH, JPG, MDi and HC collected the data. YL, OG, MDi and HC analyzed the data. YL, FDG, and HC wrote the initial draft of the paper. All the authors carefully revised the manuscript.

Supplementary material

Supplementary material for this paper can be found at the journal website: http://journals.sagepub.com/home/jcb