Abstract
Background and objective: The parkinson variant of multiple system atrophy (MSA-P) and progressive supranuclear palsy (PSP) present with atypical parkinsonism, which may be misdiagnosed as PD, particularly in early disease stages. It was previously shown that diffusion-weighted MRI (DWI) is a sensitive tool to discriminate MSA-P from PD based on increased apparent diffusion coefficients (ADCs) in the putamen. In this study DWI was evaluated in 10 patients with PSP compared with 13 patients with PD and 12 with MSA-P.
Methods: Disease was diagnosed according to established diagnostic criteria and groups were matched for age, disease duration, and Hoehn and Yahr “off” stage. Regional ADCs (rADCs) were determined in different brain regions including basal ganglia, gray matter, white matter, substantia nigra, and pons.
Results: In patients with PSP compared with those with PD, rADCs were significantly increased in putamen, globus pallidus, and caudate nucleus. Stepwise logistic regression analysis followed by receiver operating characteristics analysis identified an optimal cut-off value for putaminal rADC, discriminating PSP and PD with a sensitivity of 90% and a positive predictive value of 100%. DWI failed to discriminate PSP and MSA-P.
Conclusions: These results show that DWI detects basal ganglia abnormalities in PSP patients within few years of disease onset, discriminating patients with PSP from those with PD, but not from those with MSA-P.
The clinical differentiation of PD from atypical parkinsonian disorders (APDs) such as progressive supranuclear palsy (PSP) and the parkinson variant of multiple system atrophy (MSA-P) may be challenging, especially during the early disease stages.1-5⇓⇓⇓⇓ MRI techniques including routine, spectroscopic, and volumetric MRI, as well as diffusion-weighted imaging (DWI) have been shown to be helpful in the differential diagnosis of PD vs APD based on morphologic and functional abnormalities of the basal ganglia or brainstem.5-14⇓⇓⇓⇓⇓⇓⇓⇓⇓ An early differentiation between APD and PD is important for a number of reasons, including differences in natural history and treatment response.15-21⇓⇓⇓⇓⇓⇓ Furthermore, pharmacologic and neurosurgical treatment trials for PD require a correct diagnosis avoiding inclusion of patients with misdiagnosed APD.22-25⇓⇓⇓
In contrast to other imaging techniques such as spectroscopic and volumetric MRI, DWI is a rapid and more economical tool that is available on most of the routinely available 1.5-T MR scanners. DWI is commonly used to determine the random movement of water molecules that are aligned with fiber tracts in the CNS. Quantification of diffusion is possible by applying field gradients of different degrees of diffusion sensitization, allowing the calculation of the apparent diffusion coefficient (ADC) in tissue. Because the CNS is organized in bundles of fiber tracts, the water molecules mainly move along these structures, whereas diffusion perpendicular to the fiber tracts is restricted.26 Thus pathologic processes, such as neuronal loss and secondary astrogliosis, remove some of the “restricting” barriers, increasing the mobility of water molecules within the tissue architecture. Thus pathologic processes that modify tissue integrity can result in an increased ADC.27 In a recent diffusion-weighted MRI (DWI) study we reported that it was possible to completely distinguish patients with MSA-P from those with PD.12 Patients with MSA-P showed significantly higher regional ADC (rADC) values in the putamen than patients with PD and controls. In addition, DWI provided a full discrimination of patients with MSA-P and PD based on putaminal rADC values.
In this study we extend our previous findings by investigating the presence, distribution, and diagnostic role of DWI abnormalities in PSP, a neuronal multisystem degeneration characterized by atypical parkinsonism associated with prominent subcortical tau pathology.28,29⇓ Patients with MSA-P and PD matched for age and disease duration served as control groups.
Patients and methods.
Patients.
Twelve consecutive patients with MSA-P,30 10 consecutive patients with PSP,31 and 13 consecutive patients with PD32 matched for age, disease duration, and Hoehn & Yahr “off” stage were recruited at our parkinson outpatient clinic. Clinical diagnosis of MSA-P, PSP, and PD was made according to established criteria30-32⇓⇓ by movement disorder specialists experienced in parkinsonian disorders (G.K.W., W.P.). A detailed clinical history and a careful neurologic examination were performed. At the time of the DWI study, 10 of the patients with MSA-P had been classified as “probable,” two as “possible,” both of these being reclassified as “probable” at follow-up 1 year after DWI examination. The PSP group comprised 10 patients satisfying the National Institute of Neurological Disorders and Stroke-SPSP criteria for “probable” PSP at the time of DWI examination. Informed consent was obtained from all patients. We have also included in the current study patients whose DWI data have previously been reported (MSA-P, n = 10; PD, n = 11).12 Demographic and clinical data on the patients are listed in table 1.
Table 1 Demographic and clinical data of patients with PD, MSA-P, and PSP in this study
MRI protocol.
Conventional dual-echo fast-spin-echo and DWI sequences were performed in all patients using a 1.5-T whole-body MR scanner (Magnetom Vision, Siemens, Erlangen, Germany) and a circular polarized head coil. The dual-echo spin-echo sequence had a repetition time of 3,500 ms, an echo time of 22 and 90 ms, a slice thickness of 2 mm, a matrix of 256 × 256 pixels, and a field of view of 200 ms. This sequence was performed twice, providing 2 × 15 slices that were interleaved without any gap. DWI scans were acquired using a spin-echo type of echoplanar imaging sequence with diffusion-sensitizing gradients switched in slice direction and three different b-values (30, 300, and 1,100 s/mm2). Sequential sampling of k-space was used with an effective echo time of 123 ms, a bandwidth of 1,250 Hz/pixel, and an acquisition matrix of 128 × 128, which was interpolated to 256 × 256 during image calculation. The DWI sequence provided 20 consecutive slices with a slice thickness of 3 mm and a field of view of 230 mm. The acquisition time of each DWI sequence was 5 seconds.
Image analysis.
ADC maps were calculated by fitting the logarithm of the signal intensity as a function of the gradient factor b over three different b-values for each pixel.12 After calculation of ADC maps, rADC values were determined by segmentation of selected brain regions including putamen, caudate nucleus, globus pallidus, thalamus, substantia nigra, pons, white matter (periventricular), and gray matter (parietal cortex) using the echoplanar images with the lowest b-value. These images provide a good contrast between gray and white matter and permit a sufficient differentiation of different brain structures. The segmented regions of interests (ROIs) were copied on the ADC maps in order to obtain the mean rADC values.12 Additionally, we placed ROIs in the CSF at midventricular level) of each patient and revealed a mean rADC of 2.69 (SD 0.31) × 10−3 mm2/s.33 To avoid CSF contamination and partial volume effects, we excluded all ADC pixel values that were >2.0 × 10−3 mm2/s (corresponding to mean CSF rADC −2 SD).33,34⇓ The reliability and precision of the DWI technique including the segmentation procedure were reported previously.12
Statistical analysis.
Data were tabulated and analyzed using SPSS 10.0 for Windows (SPSS, Chicago, IL). One-way analysis of variance followed by post hoc Bonferroni correction was used for comparison of the age at examination and disease duration between groups (PD, MSA, PSP). Unified PD Rating Scale (UPDRS) “off” scores and the Hoehn & Yahr “off” stages in patients with APD and PD were compared by the Mann–Whitney U test.
Because the mean rADC values of most brain regions were not normally distributed, as revealed by the Shapiro-Wilks test, the Kruskal-Wallis test was used to compare the mean rADC values between the PD, MSA, and PSP groups. When detecting significant effects in the Kruskal-Wallis test, multiple group comparisons were performed by using post hoc Mann–Whitney U tests. rADC values and disease severity as measured by the UPDRS “off” scores were correlated for each patient group by the Spearman rank test.
To discriminate between PD and APD, forward and backward stepping logistic regression analysis was performed, including the rADCs of all ROIs, age at examination, sex, and disease duration as independent variables. Because forward and backward stepping logistic regression analysis revealed putaminal rADCs as the only significant variable to discriminate between PD and APD (see “Results” section), sensitivity and specificity values were calculated for putaminal rADCs only, using optimal cut-off values determined by receiver operating characteristics (ROC) curve analysis. This curve plots sensitivity (i.e., relative number of patients with APD identified by the abnormal rADCs) vs specificity (i.e., probability of not having APD given the absence of the abnormal rADCs) for every possible cut-off point. The positive predictive values (PPV, i.e., likelihood of a person with the abnormal rADCs having APD) and the negative predictive values (NPV, i.e., likelihood of a person without the abnormal rADCs not having APD) were calculated for the optimal cut-off value in the ROC curve.
The significance level was set at p < 0.05. Because of the multiple group comparisons, the significance level of the post hoc Mann–Whitney U tests were set at a lower threshold (p < 0.05/3 = 0.017). Spearman rank correlation coefficients of 0.35 to 0.49 were interpreted empirically as low, those of 0.5 to 0.79 as moderate, and those of ≥0.8 as high. Results are reported as means (SD) or medians (range) depending on the test used for statistical evaluation.
Results.
Patients.
Patient age was not significantly different between groups at the time of MRI examination. Mean age at examination was 62 (SD 10.5) years in PD, 63 (SD 6.6) years in MSA-P, and 68 (SD 6.9) years in patients with PSP. There were no differences in disease duration for patients with PD (3.0 years, SD 1.2), MSA-P (2.8 years, SD 1.1), and PSP (2.7 years, SD 1.1). The Hoehn & Yahr “off” stages of all three patient groups were not different (MSA-P range II to III; PD range I to III, PSP range II to III, p > 0.1). Patients with MSA-P (median 38, range 29 to 53; p = 0.001) and PSP had higher UPDRS “off” scores (median 35, range 21 to 45; p = 0.03) than patients with PD (median 26, range 13 to 38). A summary of the clinical findings is given in table 1.
Diffusion-weighted MRI.
Comparing PD, MSA-P, and PSP, the Kruskal-Wallis test revealed a difference of rADCs in the putamen (p < 0.001), in the caudate nucleus (p = 0.011), and in the globus pallidus (p = 0.028) (table 2; figure). Post hoc testing with the Mann–Whitney U test revealed an increase in putaminal rADC values in both patients with MSA-P (p < 0.001) and patients with PSP (p < 0.001) compared with patients with PD. Moreover, none of the putaminal rADC values in the PD group surpassed the lowest value in the MSA-P group. Only one patient with probable PSP showed a putaminal rADC value within the PD range. Patients with MSA-P (p = 0.007) and PSP (p = 0.015) both showed increased rADCs in the caudate nucleus compared with patients with PD as tested with post hoc Mann–Whitney U test. Further testing of the pallidal rADCs with the Mann–Whitney U test revealed an increase in patients with PSP (p = 0.005) but not MSA-P compared with the PD group. There was no significant group difference for any of the rADCs between patients with MSA-P and PSP. rADC group differences of the other ROIs showed no significance using the Kruskal-Wallis test. Spearman rank test revealed a correlation between putaminal rADC values and UPDRS “off” scores in patients with MSA (r = 0.66; p = 0.021), but not PSP or PD. The rADCs of the remaining ROIs did not correlate with UPDRS “off” scores in any patient group.
Table 2 DWI data of patients with PD, MSA-P, and PSP in this study
Figure. Scatter graph of (A) putaminal, (B) caudate, and (C) pallidal apparent diffusion coefficients (rADC) values (10−3 mm2/s) from patients with PD (group 1), patients with the parkinson variant of multiple system atrophy (MSA-P) (group 2), and patients with progressive supranuclear palsy (PSP) (group 3).
Discrimination between patients with PD, progressive supranuclear palsy, and multiple system atrophy.
By using forward and backward stepping logistic regression analysis to discriminate between PD and APD, putaminal rADCs remained the only significant variable (coefficient (β) 0.066, SE of β 0.027, OR 1.07, 95% CI for OR 1.01 to 1.12, p = 0.013). Maximal discrimination is reached at the cut-off level that has the highest sum of sensitivity and specificity. This optimal cut-off level (with an area under the curve of 0.92) for putaminal rADCs to discriminate between APD and PD was 0.760 × 10−3 mm2/s, implying that an rADC of ≥0.760 × 10−3 mm2/s is indicative of a diagnosis of APD and a rADC <0.760 × 10−3 mm2/s is indicative of a diagnosis of PD. Sensitivity for the cut-off level of 0.760 × 10−3 mm2/s was 96%, specificity 100%, the PPV 100%, and the NPV 93%. When using putaminal rADCs of ≥0.760 × 10−3 mm2/s to distinguish MSA-P from PD, optimal diagnostic accuracy for MSA-P could be obtained (sensitivity, specificity, PPV, and NPV all 100%); when using putaminal rADCs of ≥0.760 × 10−3 mm2/s to distinguish PSP from PD, a sensitivity of 90%, a specificity of 100%, a PPV of 100%, and a NPV of 93% were obtained. However, by using this cut-off level patients with PSP could not be discriminated from those with MSA-P. In general (table 3), no patient with PD was classified as having APD and no patient with MSA-P was classified as having PD, not even those two patients classified as having “possible” MSA-P at the time of the DWI examination. One of the patients with PSP was misclassified as having PD based on putaminal rADCs.
Table 3 Diagnostic classification based on DWI using a putaminal rADC cut-off level of 0.760 × 10−3 mm2/s
Discussion.
In this study, we evaluated the diagnostic role of DWI in patients with PSP using patients with MSA-P and PD matched for age and disease duration as control groups. A range of neuroimaging methods have been previously employed to differentiate PSP from other parkinsonian disorders, in particular PD or MSA-P. In studies using T2- and T1-weighted MRI in patients with PSP, several abnormal findings on routine MRI have been described.8-10⇓⇓ Based on these studies, routine MRI criteria to support a clinical diagnosis of PSP have been proposed.8 However, most of the abnormalities on routine MRI have only suboptimal diagnostic accuracy, with considerable overlap with other APDs.8,10⇓ A previous volumetric MRI study showed significant reductions in mean striatal and brainstem volumes in patients with APD including PSP and MSA compared with patients with PD and controls.11 By application of stepwise discriminant analysis, patients with APD were well discriminated from patients with PD and control subjects. However, patients with PSP could not be distinguished from patients with MSA-P.11 Previous spectroscopic studies showed reduced but overlapping n-acetylaspartate (NAA)-to-creatine and NAA-to-choline ratios in the lentiform nucleus of patients with PSP and MSA-P compared with patients with PD or controls, presumably reflecting neuronal loss.6,13,14⇓⇓ Volumetric and spectroscopic MR studies indicate neuronal loss in the basal ganglia of patients with PSP and MSA-P and may help to distinguish patients with these disorders from those with PD.6,11⇓ However, volumetric MR and spectroscopic MR studies are costly, time-consuming, and available only in specialized research centers. In contrast to MR volumetry and MR spectroscopy, DWI is available on nearly all clinical 1.5-T MR scanners. The advantages of the DWI sequence used in our study are the very short acquisition time (few seconds), the relatively high spatial resolution, and the absolute quantitation by calculating ADC maps.
In a recent study by our group, DWI completely discriminated patients with PD and MSA-P based on putaminal ADCs.12 Another DWI study found increased ADC in the prefrontal and precentral white matter in five patients with PSP compared with controls.35 However, there was no comparison with PD or other APD patients. Furthermore, basal ganglia were not analyzed. In the current study, mean rADCs in the putamen, in the caudate nucleus, and in the globus pallidus were significantly increased in patients with PSP compared with PD, whereas there were no significant differences of rADCs in the thalamus, substantia nigra, pons, periventricular white matter, and gray matter. The increased rADC values in the basal ganglia in our patients with PSP are likely to reflect ongoing neuronal degeneration and astrogliosis, whereas most neuropathologic studies reveal intact striatum in PD.36
Neuropathologically, PSP is characterized by the occurrence of neurofibrillary tangles and neuropil threads in cortical and subcortical sites, particularly in the basal ganglia and brainstem, with a variable degree of neuronal loss and gliosis due to widespread abnormal tau accumulation involving nerve cell soma and processes as well as glial cells in heavily affected areas.28,29,37-41⇓⇓⇓⇓⇓⇓ The underlying mechanism for increased water diffusion leading to an elevation in ADC may be due to neuronal loss and gliosis resulting in destruction of tissue architecture. This is in keeping with in vitro measurements that revealed an increase of the ADC due to expanded extracellular space and increased membrane permeability.26,42-45⇓⇓⇓⇓ For DWI we used a sequence with diffusion-sensitizing gradients in slice direction only, which may result in an underestimation of diffusion-related pathologic alterations in the CNS.46 This might explain the nonsignificant ADC change in other ROIs such as thalamus or brainstem regions. Possibly, the fiber tracts in these brain regions might be adversely orientated, resulting in widely scattered rADC values as indicated by our results.
The involvement of the putamen, caudate nucleus, and globus pallidus detected by DWI in patients with PSP is consistent not only with the underlying neuropathology of PSP but also with MR features including signal changes as well as atrophy in the basal ganglia,8-10⇓⇓ striatal volume loss on MR volumetry,11 and a reduced NAA-to-creatine ratio in the lentiform nucleus on MR spectroscopy.6,14⇓ However, only the volumetric study could differentiate well between patients with PD and APD (including both patients with PSP and MSA) by using an extensive image postprocessing procedure and statistical analysis. Five (83%) of the six patients with PSP included in the volumetric study11 were predicted to have APD; however, also three (27%) of the 11 patients with PD were predicted to have APD. In this study, by using stepwise logistic regression analysis followed by ROC analysis, we identified an optimal cut-off putaminal rADC value of 0.760 × 10−3 mm2/s, discriminating patients with APD and PD with a sensitivity of 96% and a PPV of 100% (see table 3). All patients with PD were classified correctly, and only one of the patients with APD (PSP) was classified as having PD based on putaminal rADCs. Corresponding to our recently published study,12 we were able to completely discriminate between a larger series of patients with MSA-P and PD based on putaminal rADCs. Furthermore, both patients with increased putaminal rADCs classified as having “possible” MSA at the time of the DWI examination were reclassified as having “probable” MSA at 1-year follow-up.
Similar to spectroscopic and volumetric MR studies,6,11⇓ we were not able to find any significant group difference between the increased rADCs of patients with PSP and MSA-P. This confirms the sensitivity of DWI in detecting degenerative basal ganglia pathology regardless of the underlying disease process, which involves abnormal aggregation of α-synuclein in MSA-P and of tau in PSP.3,28,29,47,48⇓⇓⇓⇓
Acknowledgments
Supported by the Austrian Federal Ministry of Science and Transport (GZ 70038/2 PR 4/98).
Footnotes
-
See also pages 892, 910, and 917
- Received June 10, 2002.
- Accepted November 16, 2002.
References
Letters: Rapid online correspondence
REQUIREMENTS
You must ensure that your Disclosures have been updated within the previous six months. Please go to our Submission Site to add or update your Disclosure information.
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.
