Diagnostic accuracy of Magnetic Resonance Parkinsonism Index in differentiating progressive supranuclear palsy from Parkinson's disease and controls in Indian patients
Correspondence Address: Source of Support: None, Conflict of Interest: None DOI: 10.4103/0028-3886.177611
Source of Support: None, Conflict of Interest: None
Aims and Objectives: An assessment of the sensitivity and specificity of magnetic resonance (MR) imaging measurements of midbrain, pons, middle cerebellar peduncles (MCPs), and superior cerebellar peduncles (SCPs) and MR Parkinsonism Index (MRPI) in differentiating progressive supranuclear palsy (PSP) from Parkinson's disease (PD) and controls was performed. The correlation of these MR imaging measurements with the duration and severity of disease in the Indian patients using the PSP rating scale (PSPRS) was also performed.
Keywords: Magnetic resonance imaging; Magnetic Resonance Parkinsonism Index; progressive supranuclear palsy
Progressive supranuclear palsy is a neurodegenerative disorder characterized by postural instability with falls, supranuclear vertical gaze abnormalities with Parkinsonian features, and frontal cognitive disturbances. Symmetric onset, near absence of rest tremors, predominant axial involvement, and poor response to levodopa help in differentiating progressive supranuclear palsy (PSP) from Parkinson's disease (PD). The clinical differentiation between PSP and PD is difficult in the early stages. The management and prognosis of PSP differs significantly from PD, making it essential to make an early distinction between PSP and PD. The clinical diagnostic criteria developed during a workshop at the National Institute for Neurological Disorders and Stroke, and the Society for PSP (NINDS-SPSP) in 1996 has been the most widely accepted and validated methodology for differentiating between the two entities and includes “probable” clinical criteria with a high specificity necessary for a treatment trial, “possible” criteria sufficiently sensitive for use in a prevalence study, and “definite” criteria with histopathologic confirmation.
Golbe and Ohman-Strickland proposed a clinical rating scale for PSP (the PSP rating scale [PSPRS]) to assess the disease severity. This scale also helps to predict survival. Studies have shown an association of the total score and the subscores of the PSP rating scale with structural damage of brain in PSP. The PSPRS is not a diagnostic tool, but a quantitative measure of disability. It includes all the important areas of clinical impairment in PSP. The score on this scale increases at the rate of approximately 9 points per year in patients with probable PSP.
Despite using the standard clinical criteria, a number of PSP patients are diagnosed at postmortem examination. Newer clinical phenotypes like PSP-Parkinsonism (PSP-P), characterized by asymmetric onset, early bradykinesia, tremors, late-onset falls, absent/late-onset eye abnormalities, and a good response to levodopa, do not satisfy the current clinical diagnostic criteria of PSP. In the the early stages, PSP-P may be difficult to distinguish from PD. Patients with PSP-P develop features of PSP in the late stages. Also, PSP still remains underdiagnosed due to lack of awareness and of diagnostic skills in most general physicians.
The routine magnetic resonance imaging (MRI) of the brain stem structures show midbrain and superior cerebellar peduncle (SCP) atrophy in PSP patients. These single measurements of brain stem structures fail to differentiate between PSP, PD and multiple system atrophy-P (MSA-P). The midbrain/pons (M/P) ratio was significantly smaller in PSP than in PD, MSA-P, or control patients. However M/P ratio failed to distinguish between PSP and PD on an individual basis. Magnetic Resonance Parkinsonism Index (MRPI), calculated by multiplying the pons-to-midbrain area ratio (P/M) to the middle cerebellar peduncle (MCP)-to-SCP width ratio, was introduced to circumvent this issue and has shown a high sensitivity, specificity, and diagnostic accuracy in differentiating PSP from other groups. Other diagnostic modalities that help in diagnosing PSP are the cerebrospinal fluid (CSF) studies. The CSF neurofilament content is significantly higher in PSP and MSA-P, reflecting the degree of ongoing neuronal degeneration affecting mainly the axonal compartment. The CSF levels of total tau protein in PSP were similar to those in controls. The patterns of proteolytic tau fragments in CSF in PSP were different from that seen in other neurodegenerative conditions like frontotemporal dementia, corticobasal degeneration, and PD. These are possible biomarkers for an early diagnosis of PSP, but are not accessible to an average physician.
The MRI findings in PSP are often absent in the early stages of the disease. The common MRI findings in PSP are midbrain atrophy particularly the dorsal portion, with enlargement of the third ventricle, tegmental atrophy and hyperintense signal change in the periaqueductal gray matter  indicative of gliosis, superior cerebellar peduncle atrophy, and frontal and temporal atrophy. The other described MRI signs of PSP patients are the hummingbird sign, giant penguin appearance,,, morning glory sign, and Mickey Mouse appearance,,, reduced area of the midbrain on midline sagittal images, and reduced M/P ratio [approximately 0.12 (normal, approximately 0.24] on the midline sagittal image. The accuracy of conventional MRI in predicting the neuropathological diagnosis of PSP and MSA, and the relationship between the MRI abnormalities and the macroscopic neuropathological findings have been studied. The conventional MRI predicted the diagnosis in approximately 66% of the PSP and MSA cases. Superior cerebellar peduncle atrophy and both the “hummingbird” and “morning glory” signs have a higher specificity but less sensitivity than the clinical diagnostic criteria. The diffusion-weighted imaging (DWI) shows significantly higher regional apparent diffusion coefficient (rADC) values in the putamen and globus pallidus in PSP patients, probably indicating an ongoing striatal degeneration. The volume loss of various supra- and infratentorial brain structures, measured by MR volumetry with semiautomatic segmentation techniques on a region-of-interest (ROI) approach, showed a significant reduction in the whole brain, striatal, brain stem (especially the midbrain), and frontal lobe volumes.
Voxel based morphometry (VBM) used in detecting volume loss in Parkinsonian disorders showed a gray matter loss in PSP, particularly in the frontotemporal cortical areas such as the prefrontal cortex and insular region, whereas the white matter loss was additionally reported in the midbrain, including the cerebral peduncles. Midbrain atrophy correlates with disease severity and motor deficits. Reduced anteroposterior (AP) midbrain diameter (<14 mm) and abnormal superior profile of the midbrain (flat or concave in PSP versus a convex aspect in healthy people) may assist in the differential diagnosis of PSP., Some PSP patients have increased signal changes in the SCP on fluid-attenuated inversion recovery images, a sign which is absent in PD and MSA.
The data on MRI of PSP patients in India is scarce. Our study, was designed to study the sensitivity and specificity of MRPI in differentiating PSP patients from PD patients and controls.
This study enrolled 26 probable PSP patients, diagnosed as per the criteria proposed by NINDS-SPSP, and attending the outpatient clinic of the Department of Neurology at P. D. Hinduja National Hospital, Mumbai, from March 2012 to March 2014. Thirteen patients with PD fulfilling the UK brain bank criteria were also selected. Thirty age-matched controls included individuals of age >40 years and without any brain stem pathology or clinical Parkinsonian symptoms, who attended our outpatient clinic for other neurological disorders. The study was approved by the institutional review board. Informed consent of the participating patients was obtained.
PSPRS includes 28 items divided into 6 categories: Activities of daily living, mentation, bulbar function, ocular motor function, limb motor function, and gait/midline function. The scores range from 0 to 100 with each item graded as 0-2 (six items) or 0-4 (22 items). PSP staging system defines the broad categories of motor function.
The probable PSP diagnostic criteria by NINDS-SPSP should include all five of the following clinical features: Gradually progressive disorder, onset at age 40 years or later, no evidence of competing diagnostic possibilities, vertical gaze palsy, and, slowing of vertical saccades and prominent postural instability with falls in the first year.
These include symmetric akinesia or rigidity, proximal more than distal; an abnormal neck posture, especially retrocollis; a poor response or an absence of response of the Parkinsonism manifestations to levodopa therapy; the presence of an early dysphagia and dysarthria; and, the early onset of cognitive impairment, including at least two of the following: Apathy, impairment in abstract thought, decreased verbal fluency, an imitation behavior, or frontal lobe release signs.
These include a recent history of encephalitis; the presence of alien limb syndrome, cortical sensory deficits, or focal frontal or temporoparietal atrophy; hallucinations or delusions unrelated to dopaminergic therapy; cortical dementia of the Alzheimer's type; prominent early cerebellar symptoms or prominent early and unexplained dysautonomia; severe, asymmetric Parkinsonian signs; neuroradiological evidence of relevant structural abnormalities; and, Whipple's disease, confirmed by polymerase chain reaction.
The patients with PSP included in the present study underwent assessment and staging, respectively, using the PSPRS and the PSP staging system by both the examiners independently. The demographic data including the age at presentation, sex, age at onset, and clinical features were recorded.
The PSP patients, PD patients, and age-matched controls underwent MRI of the brain using 1.5/3-T imager. The MRI examinations included T2-weighted images, T1-weighted volumetric spoiled-gradient echo images, and diffusion-weighted images. The area of the midbrain and pons were measured on midsagittal T1-weighted volumetric spoiled gradient (VSG) echo sequence. The measurement of width of the middle cerebellar peduncle (MCP) was done on sagittal T1-weighted VSG echo images, and of superior cerebellar peduncle (SCP) was measured on the T1-weighted VSG echo high-spatial-resolution oblique coronal image. Mean MCP and SCP widths were calculated by averaging values from the right and left sides. The M/P ratio was measured as the ratio of midbrain area to pons area. MRPI was calculated by multiplying pons area/midbrain area ratio by MCP width/SCP width ratio [Figure 1].
The MRPI and M/P ratio were calculated for all patients by a consultant radiologist blinded to the diagnoses of the patients. Also, similar measurements were calculated for patients with PD and controls.
To evaluate the correlation between MRPI and PSPRS, duration of disease, and the PSP staging system, Spearman correlation coefficients were calculated. To compare the M/P ratio, one-way analysis of variance was used. To assess the differences in SCP width, MCP width, pons area, and MRPI among the groups, the Kruskal–Wallis test was used. The sensitivity, specificity, positive predictive value (PPV), and diagnostic accuracy were determined for differentiating PSP from PD and controls by using the optimal cutoff values determined with receiver operating characteristic (ROC) curve analysis. To verify the agreement between the two raters, the inter-rater correlation coefficient was calculated. All tests were two-tailed, and the α level was set at P < 0.05.
Clinical PSPRS and PSP staging was used as a quantitative measure of disability. Higher scores indicated a greater morbidity.
The demographics of PSP, PD, and control patients are illustrated in [Table 1]. The mean PSPRS score by rater 1 and rater 2 were 43.81 and 43.65, respectively, with an excellent inter-rater correlation (correlation coefficient = 0.983). The mean and median PSP staging was 3.
The difference in the midbrain area in PSP versus PD patients, and in PSP patients versus the control group was statistically significant (P < 0.001); whereas, the difference of the midbrain area in the PD patients versus the controls was not significant (P = 0.31) because the midbrain area is not affected in PD. There were overlapping values of the midbrain area between the PSP and the control group. The mean SCP width in PSP was significantly lower (P < 0.05) than that in PD and controls. The difference in mean SCP width in PD and controls was not statistically significant. Therefore, the SCP width was significantly smaller in PSP but normal in PD.
The cutoff value of 12.4 using ROC curve analysis was taken to differentiate between a higher and lower MRPI. All PSP patients had higher MRPI values than PD patients and controls. All PD and control subjects (except one) had MRPI values <12 .4. There were no overlapping values of MRPI between PSP and other groups [Figure 2]a. The mean MRPI in PSP was 23.48, whereas that in PD was 9.07 and in controls was 9.45 [Table 1]. The differences in mean MRPI between PSP versus PD and PD versus controls were statistically significant (P < 0.05), but the mean MRPI of PD versus controls was not significant [Table 2]. Therefore, MRPI was significantly higher in PSP but normal in PD. Only one control had a high MRPI value, though clinically, that patient had no features of PSP or PD. MRPI was 100% sensitive, specific, and accurate in differentiating PSP from PD when a cutoff level of 12.4 was used, and MRPI was highly sensitive (96.3%), specific (100%), and accurate (98.21%) [Table 3] in differentiating PSP from controls, with a PPV and NPV of 96.15% and 96.67%, respectively. When the cutoff value of 13.55 was taken to differentiate higher and lower MRPI, as per the earlier studies, all PSP patients in our study had higher MRPI values than that seen in PD and control patients. All PD and control patients (except one) had MRPI values <13 .55. There were no overlapping values of MRPI between the PSP and other groups. MRPI with a cutoff value of 13.55 was also 100% sensitive, specific, and accurate in differentiating PSP from PD when a cutoff level of 13.55 was used, and MRPI was also highly sensitive (96.3%), specific (100%), and accurate (98.21%) in differentiating PSP from controls, with a PPV and NPV of 96.15% and 96.67%, respectively. The value of MRPI in the current study was lower as compared with the earlier studies owing to lesser duration of illness in our PSP patients. The average duration of illness in the current study was 2.56 years [Table 1]. The mean duration of PSP was 3.57 years in a study by Morelli et al., and 4.8 years in a study by Massey et al.,,
In comparison, the correlation of M/P ratio, with a cutoff value of 0.21, in PSP, PD and control patients
showed a good correlation between the control and PSP groups, as well as the PD and PSP groups (P < 0.001); no significant correlation was found in the PD and the control group. The mean M/P ratio in the PSP group was lower (0.16) than that in the PD (0.3) and control group (0.27). The difference was statistically significant (P < 0.001), but the difference in the mean M/P ratio between the PD and control groups was not significant (P = 0.052). The M/P ratio was decreased in PSP but normal in PD. The M/P ratio was 100% sensitive, 92.86% specific, and 97.44% accurate in differentiating PSP from PD when the cutoff level of 0.195 was used (PPV, 96.5%, and NPV, 100%). However, there were overlapping values in PSP and PD patients [Figure 2]b.
There was no significant correlation between MRPI and the duration of disease or between MRPI and and PSPRS using Spearman rho correlation coefficient, but there was a statistically significant correlation between the duration of disease and average PSPRS (correlation coefficient, 0.744; P < 0.001). There was no significant correlation between MRPI and the PSP staging [Table 4].
MRPI is proposed to differentiate PSP on an individual basis from PD and MSA-P. MRPI is significantly higher in PSP patients than in healthy controls, PD patients, or MSA-P patients. MRPI has a very high sensitivity, specificity, and positive predictive value when a cutoff of 12.4 was taken to differentiate these entities from PSP., In our study, PSP had a mean MRPI of 23.4 as opposed to 9.45 in controls and 9.07 in PD.
Most PSP patients have onset of disease in the seventh decade of life. The mean age at onset in our study was 63.58 years, which is comparable to that found in other studies.,, The mean age at examination in our study was 66.15 years, with the average duration of the disease being 2.56 years. Other studies have reported a higher age at examination as well as longer duration of disease. Two of our patients had an abnormal MRPI (>20) within 6 months of the disease onset. Therefore, MRPI can be helpful in differentiating PSP from PD in the early stages of the disease.,, However, further studies are required to confirm this finding.
Our study had an average PSPRS of 43.73 in comparison with 33.5 to 37 reported in other studies. This is due to the inclusion of only probable PSP patients in our study. Litvan et al., showed that the average PSPRS score at baseline was 33.5, and there was average worsening of score per year by 9 in the PSP patients.
Most patients in our study had the PSP staging score of 3. Calculation of M/P ratio can differentiate PSP from PD, MSA-P, or healthy controls., This is despite the fact that there were overlapping individual values in another study as well as this study.
Midbrain atrophy is a typical pathologic feature of PSP. In our study, the midsagittal MRI of the brain showed that midbrain area in PSP (mean, 71.88 mm 2) was almost half of the midbrain area in PD (mean, 143.15 mm 2) and controls (mean, 135.57 mm 2). This finding is consistent with the previous studies which had also shown that the midbrain area in PSP was significantly smaller than PD and controls. A midbrain area of <75 mm 2 is highly suggestive of PSP. The individual values of midbrain area measured in PSP, however, overlap with those in the control group, making the differentiation of PSP patients from other groups, on an individual basis, difficult.
In comparison with other studies, our study showed a lower mean pons area in PSP (461.9 mm 2) than in PD and controls (495.47 mm 2). There was no significant difference in the mean pons area in PSP and that in controls.,, The mean SCP width in PSP in our study was 2.34 mm, which was significantly less than that seen in PD (3.1 mm) and control (3.46 mm) subjects. Previous studies have also shown the mean SCP width to be <3 mm in patients with PSP. However, there was no significant difference in the mean SCP width in patients with PD and that in controls. Also, on an individual basis, SCP width could not differentiate the patients with PSP from other groups.,,,
As per the aforementioned data, the mean MCP width in PSP (7.85 mm) was smaller than that in PD (8.09 mm) and controls (8.62 mm), but with no significant difference. Previous studies also showed comparable results.,,,
The average values obtained in other studies include an MRPI of 21.4 by Morelli et al., 20.7 by Longeni et al., 18.63 by Hussl et al., and 24.56 by Quattratone et al., These values are comparable to our current study where the average value was 23.48. The MRPI was 9.07 in patients with PD in our study, and 9.45 in controls, a finding similar to that seen in other studies.,,,
The single brain measurements on MR images fail to aid in differentiating PSP from other groups. The combined assessment of all four brain structures involved in this neurodegenerative process overcomes this problem. The mean MRPI in PSP was 23.48, which was significantly higher than that in PD (9.07) and controls (9.45). On an individual basis, MRPI could differentiate all PSP patients from other groups, except for one control, where the person in focus had clinically no signs of Parkinsonism, but the MRPI value was >12.4.
Calculation of M/P ratio can differentiate PSP from PD, MSA-P, or healthy controls., The mean M/P ratio in PSP (0.16) was significantly lower than that in PD (0.3) and controls (0.27) in the present study. There were overlapping values between the PSP and the control groups. Other studies have also shown similar overlapping values between the PSP and other groups. There was only a marginal reduction in the sensitivity of M/P ratio in predicting PSP on an individual basis.
Thus, MRPI is 100% sensitive, specific, and accurate in differentiating PSP from PD when a cutoff level of 12.4 was used; and, MRPI is highly sensitive (96.3%), specific (100%), and accurate (98.21%) in differentiating PSP from controls. This result was comparable to that obtained in the previous studies where sensitivity, specificity, and diagnostic accuracy were all more than 90%. One possible reason for this distinct differentiation of PSP from PD obtained with the MRPI measurement may be the smaller number of patients with PD, most of whom had a longer duration of disease.
In the present study, the M/P ratio was 100% sensitive in differentiating PSP from PD but showed less specificity and diagnostic accuracy as compared with MRPI in differentiating PSP from PD and controls (although it was more than 90% sensitive). Morelli et al., also showed that M/P ratio failed to distinguish PSP from PD on an individual basis, with a relatively lower sensitivity, specificity, and diagnostic accuracy of 92.9%, 85.3%, and 86.8%, respectively. Hussl et al., also reported a very low sensitivity and diagnostic accuracy of the M/P ratio in differentiating PSP from PD., The cutoff value for MRPI in the Indian patients with PSP was different from that seen in the earlier studies.,
Previous studies have reported a significant correlation between MRPI and the duration of disease, but we found no correlation of MRPI with either PSPRS or the duration of disease , [Table 4]. This discrepancy may be due to the shorter duration of disease in some of our patients. Litvan et al., have shown worsening of total PSPRS by an average of 9.1 per year. There was significant correlation between PSPRS and the duration of disease in our study. We, however, did not find any correlation between MRPI and the PSP staging system.
MRPI improved the diagnostic accuracy much more than that seen with the use of M/P ratio in discriminating PSP from MSA-P because of the combined assessment of four brain structures involved in neurodegenerative disorders. Thus, atrophy of midbrain and SCP suggested the presence of PSP, and atrophy of pons and MCP suggested the presence of MSA. These brain structures are usually not involved in PD. However, calculation of MRPI is more time consuming than calculation of the M/P ratio because it involves measurement of additional brain structures. The limitation of our study was the lack of pathological confirmation of PSP and that patients with atypical presentations of PSP were not included in this study.
Magnetic Resonance Parkinsonism Index is more sensitive, specific, and accurate in differentiating PSP from PD in the early stages on an individual basis. This index may be an important diagnostic tool available to a physician for the diagnosis of PSP.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
[Figure 1], [Figure 2]
[Table 1], [Table 2], [Table 3], [Table 4]