Abstract
We assessed the validity and interrater reliability of neurologists who, using four different sets of previously published criteria for the clinical diagnosis of progressive supranuclear palsy (PSP), also called Steele-Richardson-Olszewski syndrome, rated 105 autopsy-proven cases of PSP (n equals 24), Lewy body disease (n equals 29), corticobasal ganglionic degeneration (n equals 10), postencephalitic parkinsonism (n equals 7), multiple system atrophy (n equals 16), Pick's disease (n equals 7), and other parkinsonian or dementia disorders (n equals 12). Cases were presented in random order to six neurologists. Information from each patient's first and last visits to the medical center supplying the case was presented sequentially to the rater, and the rater's diagnosis was compared with the neuropathologic diagnosis of each case. Interrater agreement for the diagnosis of PSP varied from substantial to near perfect, but none of the criteria had both high sensitivity and high predictive value. Because of these limitations, we used a logistic regression analysis to identify the variables from the data set that would best predict the diagnosis. This analysis identified vertical supranuclear palsy with downward gaze abnormalities and postural instability with unexplained falls as the best features for predicting the diagnosis. From the results of the regression analysis and the addition of exclusionary features, we propose optimal criteria for the clinical diagnosis of PSP.
NEUROLOGY 1996;46: 922-930.
TX.-
In 1964, Steele, Richardson, and Olszewski described progressive supranuclear palsy (PSP), a brain neuro-degenerative disorder characterized by postural instability, parkinsonism, vertical supranuclear palsy, pseudobulbar palsy, and mild dementia. [1,2] However, the first clinical report on PSP [3,4] (also called Steele-Richardson-Olszewski syndrome) was published in 1904 and was soon followed by others. [5-7] When fully expressed, the clinical signs and symptoms of PSP are usually reliable for making the correct diagnosis, but cardinal signs, such as ophthalmoplegia, may be absent, or the only symptoms may be dementia or akinesia. [7-13] Neuropathologic examination remains the ``gold standard'' for the diagnosis of PSP. A clinical diagnosis can be mistakenly applied to pathologically determined diffuse Lewy body disease, corticobasal ganglionic degeneration (CBGD), cerebrovascular disease, Pick's disease, or subcortical gliosis. [14-20] Conversely, neuropathologically confirmed PSP may be clinically mistaken for idiopathic Parkinson's disease (PD), pallidonigroluysial atrophy, cerebrovascular disease, Alzheimer's disease (AD), or CBGD. [20-26]
In attempts to improve the accuracy of its clinical identification, investigators have proposed numerous diagnostic criteria for PSP, [27-32] most of which have mandatory requirements, [27,29-32] although the features, both mandatory and nonmandatory, vary. The present study focused on the reliability and validity of four different sets of clinical criteria, as well as the raters' clinical judgment, for the diagnosis of PSP. Our goal was to identify criteria that would provide high sensitivity and high positive predictive value for the diagnosis of PSP at an early stage so that patients could be recruited for clinical studies.
Methods
Sample and data collection
Cases were selected from the research and clinical files of seven medical centers by neuropathologists who used for their diagnoses the recently published National Institute of Neurological Disorders and Stroke neuropathologic criteria for the diagnosis of PSP and related disorders, [20] the reliability of which is substantial, [33] and Kosaka's proposed neuropathologic criteria for Lewy body disease. [34] There were 105 cases with a neuropathologic diagnosis of PSP (n equals 24), Lewy body disease (n equals 29, comprising 15 cases of PD and 14 cases of diffuse Lewy body disease), CBGD (n equals 10), multiple system atrophy (n equals 16), postencephalitic parkinsonism (PEP) (n equals 7), Pick's disease (n equals 7), Creutzfeldt-Jakob disease (n equals 4), AD with extrapyramidal features (n equals 4), multiinfarct disease with parkinsonism (n equals 3), and Whipple's disease (n equals 1). These diseases were chosen because of the difficulties often presented in their clinical differentiation. All cases met the criteria for inclusion in the study (i.e., neuropathologic diagnosis with at least a 75% certainty, complete neurologic examinations on the first and last clinical visits, data missing for only two of the major features of the four criteria, and a neuroocular motor examination).
Demographic characteristics of the patients are shown in Table 1. Patients with multiple system atrophy, Lewy body disease (in particular, PD), and PEP had a significantly earlier age of onset, and patients with the last two disorders also survived significantly longer than patients with the other disorders (p less than 0.005).
Table 1. Demographic characteristics of the 105 patients
The case records were abstracted on standardized forms by eight of us (M.V., R.C., R.P., A.M., K.J., L.D., I.L., and C.M.), who followed strict instructions, such as to record as missing any features that were not explicitly described in the records and to record clinical descriptions uniformly, according to specific definitions provided. Cases were summarized by two of us (I.L. and L.D.) and included demographic data, age at onset of symptoms, disease progression, medications, family history of neurodegenerative disorders, associated disorders, risk factors, results of physical and neurologic examinations, and results of clinical laboratory studies. The clinical vignettes were presented in random order to six raters (three senior neurologists, who are movement disorder experts, and three junior neurologists) who were unfamiliar with the cases and the design of the study. Information from the patient's first and last visits to the medical center providing the case was presented sequentially to the same neurologist. The clinical vignette for the last visit included all the clinical, laboratory, and neuroimaging information available between the first and last visits. After reviewing the information from the first visit, the raters completed a standardized form and formulated a diagnosis first from their clinical judgement alone and then from the four different sets of criteria for the clinical diagnosis of PSP (see Diagnostic criteria). For the first visit, the laboratory and neuroimaging data were not used for making the diagnosis. The same procedure was followed for the last visit, except that the available laboratory and neuroimaging data were also considered.
Diagnostic criteria.
The four sets of clinical criteria for the diagnosis of PSP, which differ in their requirements Table 2, were those proposed by Lees [27] (criteria 1), Golbe and Davis [30] (criteria 2), and Blin et al. [28] (criteria 3 and 4). Criteria 1 requires the presence of vertical supranuclear palsy with downward gaze abnormalities in a middle-aged individual with no familial history of neurologic disease and at least two of five other features characteristic of PSP. Criteria 2 specifies, in addition, the presence of bradykinesia and has features that exclude multiple system atrophy. [35] Criteria 3 and 4 yield a diagnosis of ``probable'' PSP (criteria 3) or ``possible'' PSP (criteria 4), depending on the number of features present. Criteria 3 has nine features, all of which must be met, whereas in criteria 4, at least seven of the nine features must be present. The cases diagnosed by criteria 3 (probable PSP) overlap with those diagnosed by criteria 4 (possible PSP).
Table 2. Clinical criteria for the diagnosis of PSP
Overall, the six raters made 630 observations on the 105 cases (105 cases times 6 raters). Data were missing for less than 10% of the features in each criteria, with the exception of two features. One of these was poor or no response to levodopa therapy, specified in criteria 2, 3, and 4, where data were missing in 318 observations (53 cases times 6 raters). The other was the neck in posture of extension, specified in criteria 2, where data were missing in 72 observations (12 cases times 6 raters).
Statistical analysis.
The accuracy of the criteria for the clinical diagnosis of PSP was tested by various measures of reliability and validity.
Reliability measures
Reliability was measured by the generalized kappa statistic. Like a correlation coefficient, kappa varies from minus 1.0 (complete disagreement) to 0 (chance agreement) to plus 1.0 (perfect agreement). Strength of agreement was designated poor (kappa less than 0), slight (kappa equals 0 to 0.20), fair (kappa equals 0.21 to 0.4), moderate (kappa equals 0.41 to 0.60), substantial (kappa equals 0.61 to 0.80), and near-perfect to perfect (kappa equals 0.81 to 1.0), as previously suggested. [36] The kappa statistic was chosen as an outcome measure of agreement because it takes chance into account and, therefore, is markedly superior to a percentage-of-agreement measure.
Validity measures.
The raters' diagnoses were compared with the neuropathologic diagnoses in terms of sensitivity, specificity, and positive and negative predictive values (see Table 6 for definitions). A sample with a low frequency of PSP was intentionally chosen to mirror closely what neurologists experience in real life. Specificity could be artificially high secondary to the high frequency of non-PSP cases. Therefore, sensitivity and positive predictive value were chosen as the outcome measures, as they could more faithfully indicate the validity of a criteria for the diagnosis of PSP. The results of the other measures are shown so that they can be compared with accuracy measures used in other studies Table 3.
Table 6. Definitions of terms used in the statistical analysis.
Table 3. Reliability and validity of the clinical criteria for the diagnosis of PSP
Comparison of accuracy measures.
The pooled kappa test was used to determine the significance between kappa values. [37] The Cochran Q test was used to evaluate differences in sensitivity and specificity between the raters and between the criteria. [38] The McNemar test was used to evaluate the differences in sensitivity and specificity between the first visit and last visit. [39] Spearman's rho test was used to measure the degree of correlation between the different criteria, [40] with criteria 3 and 4 being considered as one criteria. Significance was defined as p less than 0.05.
Analysis of alternative criteria.
A stepwise logistic regression analysis [40] was performed on a calibration sample (75% of the cases randomly chosen from the whole data set) to identify the variables that would best predict the diagnosis of PSP. A separate analysis was performed for each of the six raters. The variables used in the analysis were most of the main features of the four criteria Table 2, with the exclusion of duplicate features (e.g., progressive disease course). The validity of the best predictive variables was tested in an independent validation sample (the remaining 25% of the cases). The results of this analysis were used to modify the existing criteria.
Results.
Reliability.
When the raters used the four criteria, strength of agreement for a diagnosis of PSP varied from substantial to near perfect for the first visit and was substantial for the last visit Table 3. When the raters used their clinical judgement, the interrater reliability for the diagnosis of PSP was also substantial Table 3. The differences in reliability across criteria were significant at the first and last visits (p less than 0.001), and criteria 1 was the most reliable. However, at the last visit, the reliability of the raters' clinical judgement was significantly better than that of criteria 1 (p less than 0.001).
Validity
Criteria 1.
Mean sensitivity of criteria 1 (Lees [27]) was low for the first visit (53%) but improved for the last visit (78%), and mean positive predictive values for each visit were relatively low (77% and 65%) Table 3. The sensitivity of the criteria was increased significantly between the first and last visits for only three of the six raters (p less than 0.05, McNemar test). For the last visit, the specificity was decreased significantly for only two of the raters (p less than 0.005, McNemar test). For the first visit, interrater differences in the sensitivity and specificity were not significant, but for the last visit these differences were significant (both p less than 0.01, Cochran Q test). There were no significant differences in the sensitivity and specificity of the criteria between junior and senior neurologists.
There were 23 false-positive diagnoses for the first visit, with 12 in CBGD, 6 in Lewy body disease, 3 in multisystem atrophy, and 1 each in multiinfarct disease and PEP. For the last visit, there were 61 false-positive diagnoses, with 17 in CBGD, 10 each in Lewy body disease and multiple system atrophy, 7 in PEP, 6 each in Creutzfeldt-Jakob disease and Whipple's disease, 4 in multiinfarct disease, and 1 in Pick's disease.
Criteria 2.
For the first and last visits, mean sensitivity of criteria 2 (Golbe and Davis *RF 30
*) was low (49% and 67%) and mean positive predictive values for each visit were substantial (85% and 76%) Table 3. Between the first and last visits, the sensitivity of criteria 2 was increased significantly for only two of the six raters (p less than 0.05, McNemar test), but the specificity was unchanged. Interrater differences in the sensitivity were significant for the last visit only (p less than 0.001, Cochran Q test), but differences in specificity were not significant. Differences in the sensitivity and specificity between junior and senior neurologists were not significant.
There were 12 false-positive diagnoses for the first visit, with 7 in CBGD, 4 in Lewy body disease, and 1 in multi-system atrophy. For the last visit, there were 30 false-positive diagnoses, with 12 in CBGD, 10 in Lewy body disease, 5 in multisystem atrophy, and 1 in AD.
Criteria 3.
Mean sensitivity of criteria 3 (Blin et al., *RF 28
* probable PSP), which required the presence of all nine typical features of PSP, was very low for the first visit (13%) and, although somewhat improved, remained low for the last visit (34%). In contrast, mean positive predictive values were nearly perfect for each visit (100% and 85%) Table 3. The sensitivity was significantly different between the first and last visits for only two of the six raters (p less than 0.05, McNemar test), but the specificity did not change between visits. Interrater differences in the sensitivity and specificity were not significant.
The raters made no false-positive diagnoses for the first visit. For the last visit, there were nine false-positive diagnoses, with four in CBGD, three in multiple system atrophy, and two in Lewy body disease.
Criteria 4.
Mean sensitivity of criteria 4 (Blin et al., [28] possible PSP) was low for the first visit (55%) but was improved considerably for the last visit (89%). However, the mean positive predictive values were low for each visit (73% and 50%) Table 3. The sensitivity was increased significantly between the first and last visits (p less than 0.005, McNemar test) for all but one of the six raters. Conversely, for the last visit, the specificity was decreased significantly for all six raters (p less than 0.005, McNemar test). Interrater differences in sensitivity were not significant, but differences in specificity were significant for the last visit only (p less than 0.03, Cochran Q test). Differences in specificity and sensitivity between junior and senior neurologists were not significant.
A total of 29 false-positive diagnoses were made for the first visit, with 11 in CBGD, 9 in multiple system atrophy, 8 in Lewy body disease, and 1 in Creutzfeldt-Jakob disease. For the last visit, the raters made 129 false-positive diagnoses, which is nearly the same as the number of correct diagnoses. There were 48 false-positive diagnoses in multiple system atrophy, 26 in Lewy body disease, 19 in CBGD, 10 in PEP, 9 in Creutzfeldt-Jakob disease, 8 in AD, 4 each in Whipple's disease and Pick's disease, and 1 in multiinfarct disease.
Clinical judgement.
When the diagnosis of PSP was based on the raters' clinical judgement, mean sensitivity for the first visit (72%) was higher than it was for any of the four criteria but the mean positive predictive value was low (76%) Table 3. For the last visit, sensitivity was 80%, and the positive predictive value was unchanged.
False-positive misdiagnosis for the first and last visits occurred mainly with cases of CBGD, multiple system atrophy, and, rarely, Lewy body disease. Pick's disease, Whipple's disease, and Creutzfeldt-Jakob disease. False-negative misdiagnosis occurred mainly with Lewy body disease (idiopathic PD for the first visit and diffuse Lewy body disease for the last visit), multiinfarct disease with parkinsonism, multiple system atrophy, and, rarely, AD but did not occur with CBGD. Differences in the sensitivity and specificity of clinical judgment for the diagnosis of PSP between the first and last visits were not significant.
Comparison of criteria.
Differences in sensitivity and specificity of the four different criteria, analyzed by rater for the first and last visits, were significant (all p less than 0.005, Cochran Q test). For each rater, criteria 3 was less sensitive and more specific than the other criteria. Sensitivity for the diagnosis of PSP for the first visit was significantly better when the raters used their clinical judgment than when they used any of the four criteria (p less than 0.0001, Cochran Q test). Conversely, the specificity was similar or significantly worse when they used their clinical judgement (p less than 0.001, Cochran Q test). Correlations between the criteria were highly significant for both the first (r equals 0.71 to 0.80) and last visits (r equals 0.54 to 0.71) (both p less than 0.001, Spearman's rho test), suggesting a strong relation between the diagnostic criteria. Clinical judgement was also highly correlated with all the criteria (r equals 0.47 to 0.57, p less than 0.001, Spearman's rho test).
Alternative criteria.
Because none of the existing criteria had high sensitivity and positive predictive values, we searched for alternative criteria that would improve the diagnosis of PSP.
Best predictive variables.
Logistic regression analysis on the calibration sample revealed, for each rater, on both the first and last visits, two best predictive variables: vertical supranuclear palsy with downward gaze abnormalities and severe postural instability with unexplained falls. Sensitivity ranged from 59 to 68% for the first visit and 56 to 88% for the last visit. Positive predictive values ranged from 75 to 85% for the first visit and 65 to 100% for the last visit. In addition to these two variables, inefficacy of levodopa therapy was an important predictor of PSP for some raters.
Testing of the validity of the two best predictive variables the independent validation sample showed that sensitivity ranged from 33 to 100% for both visits and positive predictive values ranged from 50 to 83% for the first visit and 28 to 100% for the last visit. Therefore, the diagnosis of PSP was not improved by using as diagnostic criteria only the two best predictive variables.
Fully met and modified criteria.
Criteria 1, 2, and 4 had rather low sensitivity or included a considerable number of false-positive cases, but criteria 3, which demands the fulfillment of all the features for a diagnosis of PSP, had a perfect (100%) positive predictive value. Therefore, we used the data set to test whether cases having all the features specified in criteria 1 and 2 would improve their validity. When all the features of criteria 1 were met, the sensitivity and positive predictive value did not improve to the extent of those of criteria 3 Table 4. Although the positive predictive value of criteria 2 markedly improved, the sensitivity severely decreased, and, for the first visit, many raters were unable to detect a single case of PSP.
Table 4. Alternative criteria for the clinical diagnosis of PSP
We then modified criteria 1, 2, and 4 to include as mandatory vertical supranuclear palsy with downward gaze abnormalities and severe postural instability with unexplained falls, the best predictive variables identified by the logistic regression analysis. With the inclusion of these mandatory features, the positive predictive value of criteria 4 improved without serious worsening of the sensitivity, but for criteria 1, the improvement was not as marked Table 4. Criteria 2 was not improved.
Optimal criteria.
We developed, and tested on the whole data set, optimal criteria for the diagnosis of PSP by using mandatory inclusionary and exclusionary criteria to rule out other parkinsonian disorders Table 5. For the first visit, the mean sensitivity was 57% (range 50 to 63%), which was better than that of any of the four existing criteria, and the mean positive predictive value was 85% (range 77 to 93%). For the last visit, the mean sensitivity was 66% (range 39 to 79%), and the positive predictive value was similar to that of criteria 2 (mean 76%, range 69 to 83%). Although the sensitivity of the optimal criteria was lower than that obtained with clinical judgment, the positive predictive value was markedly better for the first visit. Thus, for the first visit, the sensitivity and positive predictive value of the optimal criteria were superior to those of the existing criteria and the raters' clinical judgement for the diagnosis of PSP.
Table 5. Optimal criteria for the clinical diagnosis of PSP
Discussion
Reliability
The four criteria tested were very reliable for the diagnosis of PSP. There was also substantial agreement when the raters used their clinical judgement to make the diagnosis. Criteria 1 had the highest interrater reliability at the early and late stages of the disease. Variability may be explained by raters' differences in interpretation of clinical phenomena and diagnostic rules. This was more obvious when clinical information lacked sufficient details. The findings of our study are difficult to compare with those of other reliability studies, [41-43] because most of those studies usually contrasted disease cases with normal cases. However, the interrater agreement in our study (kappa equals 0.65 to 0.81) is, in general, higher than that found in other studies evaluating the widely used National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer's and Related Disorders Association (NINCDS-ADRDA) criteria [44] for the diagnosis of AD. Most of those studies reported moderate agreement for differentiating probable or possible cases of AD from non-AD cases: (kappa equals 0.64), [41] (kappa equals 0.36, probable; kappa equals 0.65, possible), [42] (kappa equals 0.38, probable; kappa equals 0.43, possible). [43] In the present study, agreement was also higher (kappa equals 0.73) in differentiating cases of non-PSP from cases of probable and possible PSP (criteria 3 and 4). Nevertheless, good interrater agreement does not mean that raters achieve the correct diagnosis; in fact, raters can all agree but be wrong.
Validity
The four criteria tested had a very low to low sensitivity (13 to 55%) for the diagnosis of PSP for the first visit, which, on average, occurred more than 3.5 years after the onset of PSP symptoms. Even though, as expected, sensitivity for the diagnosis of PSP improved for the last visit, which, on average, occurred more than 2 years after the first visit, it remained low (34 to 78%) for all but criteria 4 (89%). Our data support the notion that PSP is underdiagnosed, [45-47] but contrary to the generally accepted belief, the diagnosis of PSP is not very accurate. Positive predictive values were low (50 to 85%) for all but criteria 3 (85 to 100%). Thus, unless all the typical featres of PSP in criteria 3 were present, the misdiagnosis of PSP was high. The presence of all the features of a criteria did not necessarily improve the diagnosis. When all the features of criteria 1 were present, the sensitivity was still lower than that of criteria 3 and the positive predictive value was not as high. When all the features of criteria 2 were present, the sensitivity was extremely low but the positive predictive value was markedly improved. Consequently, criteria 3 is the most accurate of the four criteria, which is compatible with the intent for which it was developed: the recruitment of patients for research studies. [28] Because of its low sensitivity, however, its use would severely limit the number of patients that could be recruited. Criteria 2, which had a higher sensitivity but a lower positive predictive value than criteria 3, might be a good compromise, as it would be able to detect more cases but would be less accurate. With available routine neurologic information, it may be difficult to improve the sensitivity of criteria for the diagnosis of PSP without decreasing the positive predictive value. For example, for the first visit, when the raters used their clinical judgement, the sensitivity was higher than it was in criteria 2 but the positive predictive value was lower.
The low frequency of PSP in our sample, intentionally chosen to reflect the rarity of this disorder, makes it difficult to compare the study's validity with that of other studies [48,49] because most other studies had a high frequency of the target disease, which artificially inflates the specificity of diagnostic criteria. Moreover, in most of those studies, the nontarget disease cases were normal control subjects instead of disorders that may be difficult to differentiate clinically. Because those studies usually evaluated cases at end stages of the disease, their results will be cautiously compared with ours for the last visit. Studies using the NINCDS-ADRDA criteria for distinguishing AD from different types of dementias [48,49] reported a higher sensitivity (90% and 92%) than ours (range 34 to 89%), but the specificity (23% and 65%) was much lower than ours (range 74 to 98%). In the AD study, [49] the positive predictive values were lower in possible disease cases (63%) but higher in probable disease cases (91%), as would be expected and as we found in our study.
The high sensitivity of the raters' clinical judgement and the relatively low positive predictive values suggest that the raters may have been biased toward making this particular diagnosis because they had to use the existing criteria for the diagnosis of PSP. On the other hand, as there was a wide range of raters' responses, some of them may have taken into consideration more features than the ones explicitly provided in the criteria (e.g., a history of encephalitis lethargica or oculogyric crisis). Clinical judgement may have been markedly enhanced if the neurologists could have actually interviewed and examined the patients.
Misdiagnosis
Misdiagnosis varied according to the criteria used, but, in general, PSP was confused with CBGD, multiple system atrophy, and diffuse Lewy body disease. Indeed, all the raters misdiagnosed two cases of CBGD, one of which has been previously described. [19] The course and symptomatology in these cases were similar to those of PSP, including the presence of vertical supranuclear palsy with downward gaze abnormalities and postural instability with unexplained falls. However, these cases did not include the typical alien hand syndrome, unilateral dystonia, myoclonus, aphasia, or limb apraxia usually found in CBGD. Finally, even the neuropathologic differentiation of PSP and some cases of CBGD is difficult. [33] Thus, it may not be surprising that these two disorders would, in some cases, be difficult to differentiate.
Misdiagnosis occurred more frequently with criteria 4 than with criteria 1 because the latter includes vertical supranuclear palsy with downward gaze abnormalities as a mandatory feature and can rule out most cases of multiple system atrophy, in which downward gaze ophthalmoplegia is usually absent. There are fewer chances for misdiagnosis with criteria 2, which has mandatory features focusing on the exclusion of multiple system atrophy cases (i.e., early or prominent cerebellar signs and noniatrogenic dysautonomia). All these criteria need more features for excluding cases of CBGD and diffuse Lewy body disease. In fact, this was our rationale for developing the proposed optimal criteria.
Proposed optimal criteria
The proposed optimal criteria attempts to separate PSP from some of the major related parkinsonian disorders by excluding cases of PEP (no history of encephalitis), CBGD (alien hand syndrome, unilateral dystonia), multiple system atrophy (early and prominent cerebellar signs, noniatrogenic dysautonomia), diffuse Lewy body disease (hallucinations, early cortical dementia), and multiinfarct disease with parkinsonism (a focal lesion on neurologic examination or CT). Although our study did not consider the contribution of MRI, the presence of unidentified bright objects on the MRI should not be considered focal lesions. *RF 50,51
*
The proposed optimal criteria had better sensitivity and positive predictive value than the ones tested in this study. It fulfilled our main objective of identifying a criteria for research purposes that would have a high positive predictive value without a considerable decrease in sensitivity at an early stage. Because PSP has a low prevalence, [45] a low sensitivity would make it harder to identify cases for therapeutic trials or epidemiologic studies. On the other hand, in therapeutic trials, the chances of obtaining statistically significant results depend more on population homogeneity than on highly sensitive criteria for case detection. The accuracy of the proposed optimal criteria is far from desirable, as less than two-thirds of the PSP cases were identified, and the positive predictive values were 85% for the first visit and 76% for the last visit.
Our study suggests that vertical supranuclear palsy with downward gaze abnormalities and severe postural instability with unexplained falls are the main features characterizing PSP, and their addition to criteria 1 and 4 as mandatory features improved the positive predictive value without a marked decrease in sensitivity.
The findings also suggest that the ability of clinical information to improve the diagnostic criteria is limited. The proposed optimal criteria was no better than the ``possible'' criteria of Tolosa et al., [31] which only attempts to exclude cases of CBGD, multiple system atrophy, and multiinfarct disease with parkinsonism.
Study limitations
The present study has some limitations, but they do not invalidate the conclusions. For instance, autopsy studies, from which our population was drawn, may include more atypical patients than those who are usually seen in medical practice. At the present time, however, autopsy studies are the only ones that allow certainty of the diagnosis. In our study, the large sample size--the largest published in these types of studies--may cause averaging out of atypical cases. Validity was higher (100%) in a prospective study of AD cases, [52] selected according to the NINCDS-ADRDA criteria, [44] than in prospective studies of unselected cases of clinically diagnosed idiopathic PD (76%). [25,53] Although in the present study multicenter collaboration and retrospective collection of data were needed to obtain the required sample size, the seven medical centers providing the cases are specialized in evaluating these types of disorders. We could also assume that using these criteria in prospective studies would achieve better results.
The clinical vignettes may not be equivalent to the clinical examination of a patient, but they were thought to be complete by all the study participants and, therefore, may accurately reflect the difficulty of the diagnosis of PSP using the existing criteria. Furthermore, it is unclear why there were no significant differences in the diagnosis of PSP between senior and junior neurologists. One explanation may be that the junior neurologists were trained at well-recognized institutions devoted to the study of movement disorders. Another possibility is that the use of clinical vignettes ensured that all raters had exactly the same information on each case. Therefore, the seniors' experience may be more relevant to clinical examination.
It may also be questioned whether the difficulty in the diagnosis of PSP is secondary to limited clinical abilities, retrospectively collected data, or the biology of the disorder. As the disease progressed, PSP was easier to identify (high sensitivity) but more difficult to differentiate from the other disorders (low positive predictive value). The implication of this finding for the biology of the disorder is still unclear. The convergence of clinical symptomatology may correspond to the presence of lesions in similar topography. Alternatively, some of these disorders may be different phenotypic presentations of the same clinicopathologic entity.
Finally, although data on the effectiveness of levodopa therapy were missing in many of our cases, dopaminergic therapy is often not administered when parkinsonian symptoms are absent (e.g., Pick's disease) or not troublesome.
Improving the accuracy of PSP diagnosis.
Neuropsychologic or electrooculographic testing, sleep studies, P-300 brain evoked potentials, PET studies with deoxyglucose metabolism, and D2 dopamine receptor uptake support the diagnosis of PSP. [54-58] Unfortunately, the absence of neuropathologic confirmation by these tests does not allow us to state conclusively that their use would improve the accuracy of the PSP diagnosis.
The proposed optimal criteria, developed in response to the obvious failure of the existing criteria to exclude other disorders, is not perfect but, in terms of both sensitivity and positive predictive value, may be the most useful for selecting a homogeneous population of patients with PSP. The proposed optimal criteria need to be validated in studies using prospectively collected data. Prospective studies may also clarify the role of certain features, such as the ineffectiveness of levodopa therapy, as predictors for the diagnosis of PSP and may identify other signs or symptoms of the disease. Meanwhile, the proposed optimal criteria may enable us to gather a homogeneous population that could reveal biologic markers that will yield an earlier or more precise diagnosis of PSP or help us to find an effective therapy for this devastating disorder.
Acknowledgments
We thank Drs. A. Lees, L. Golbe, G. Roman, K.B. Nelson, and D.W. Anderson for helpful suggestions in the design of the study; Ms. M. Eaglin and Ms. D. Schoenberg for support in the design of the standardized forms; and Ms. B.J. Hessie for skillful editing. Drs. J.J. Hauw, P. Lantos, S. Daniel, D.W. Dickson, and D. Horoupian also provided some of the cases and some neuropathologic diagnoses.
- Copyright 1996 by Advanstar Communications Inc.
REFERENCES
Letters: Rapid online correspondence
REQUIREMENTS
If you are uploading a letter concerning an article:
You must have updated your disclosures within six months: http://submit.neurology.org
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.
You May Also be Interested in
Hemiplegic Migraine Associated With PRRT2 Variations A Clinical and Genetic Study
Dr. Robert Shapiro and Dr. Amynah Pradhan