What are the obstacles for an accurate clinical diagnosis of Pick's disease? A clinicopathologic study

More than 100 years ago, Arnold Pick described several patients who presented with progressive behavioral changes (apragmatism, outbursts of rage, and in later stages, mutism) and who, at autopsy, had characteristic frontal or temporal lobar atrophy.^1,2 Alois Alzheimer histologically characterized the disorder when he described"argentophylic globes" in the cytoplasm of neurons and the presence of ballooned neurons and spongy cortical wasting in the absence of neurofibrillary tangles or plaques.³ Pick's disease(hereafter referred to as Pick's) is still considered a relatively rare neurodegenerative disorder, affecting subjects in their 60s with the progressive development of frontal lobe type features (e.g., difficulty planning, reasoning, abnormal social behavior), language disturbances(decreased fluency followed by echolalia, mutism), later followed by memory and gait abnormalities and occasional parkinsonism.^4-8 There are no markers for the clinical diagnosis of the disease; neuropathologic examination is the gold standard for its definitive diagnosis.^9,10 Pick's is characterized by severe atrophy, neuronal loss and gliosis in the frontotemporal lobes, swollen neurons (Pick cells), and by characteristic argentophylic neuronal inclusions (Pick bodies).^10-13 Whether frontal degeneration (frontal lobe dementia) corresponds to Pick's remains controversial.^8,13-17 Neuroscientists argue whether frontal degeneration without Pick bodies is sufficient for the diagnosis of Pick's and whether Pick's is a nosological entity or a syndrome complex. When specific criteria for the neuropathologic diagnosis, including the presence of Pick bodies, are used, near-perfect agreement in diagnosing Pick's disease is achieved.¹⁸

Because frontal lobe-type features occur in other disorders (e.g., Alzheimer's disease [AD], progressive supranuclear palsy [PSP], corticobasal degeneration [CBD], and even in Parkinson's disease), it is not surprising that Pick's is often confused with other neurodegenerative disorders. In fact, Mendez et al.¹⁹ evaluated a consecutive series of autopsy-confirmed Pick's cases for clinicopathologic correlations and found that only 14% of patients had an accurate diagnosis before death, which suggests that misdiagnosis in life is common.

The challenge for the clinician who is confronted with a patient with atypical dementia is whether to define it or merely to designate it as"atypical." In the latter case, the prognosis and course of the disease will be unknown to both the clinician and the patient. An accurate clinical definition is relevant not only for clinical care but also for genetic, pharmacologic, and epidemiologic studies because firm conclusions can only be achieved from well-defined populations. Several studies have evaluated the accuracy of the clinical diagnosis of AD,^20-29 but only a few have investigated the accuracy of the clinical diagnosis of other types of neurodegenerative dementias.^30-33 In the present study, we investigated the ability of neurologists to make an accurate diagnosis of Pick's, the diagnostic pitfalls, and early predictors of autopsy-confirmed cases of Pick's.

Methods. Sample. Using recently defined neuropathologic criteria of the National Institute of Neurological Disorders and Stroke(NINDS)¹¹ and Kosaka's criteria,³⁴ we chose for study 105 cases of neuropathologically confirmed Pick's (n = 7) and non-Pick's (n = 98), including CBD (n = 10), PSP (n = 24), diffuse Lewy body disease (n = 14, currently called Lewy body dementia³³), multiple system atrophy (n = 16), Parkinson's disease (n = 15), postencephalitic parkinsonism (n = 7), Creutzfeldt-Jacob disease (n = 4), AD(n = 4), multi-infarct encephalopathy (n = 3), and Whipple's disease (n = 1). The criteria for Pick's include (1) the presence of lobar atrophy of the frontal and anterior temporal lobes (with or without atrophy of the caudate nucleus, pallidum, or substantia nigra); (2) Pick argyrophilic inclusions(Pick bodies); (3) Pick cells; (4) massive neuronal loss, astrogliosis, spongiosis (often laminar); and (5) exclusion of large or numerous infarcts, neurofibrillary tangles, Lewy bodies, changes diagnostic of AD, oligodendroglial argyrophilic inclusions, and prion P-positive amyloid plaques. Cases were selected from research and clinical files by experienced neuropathologists who had previously achieved near-perfect agreement using the NINDS criteria for Pick's.¹⁸ In addition, patients had to have complete neurologic examinations for the first and last clinic visits by "primary" neurologists. The same sample was previously used in a study of the accuracy of clinical criteria for the diagnosis of PSP and CBD.^31,32 As previously reported,^31,32 the case records were abstracted on standardized forms by neurologists who were unaware of the neuropathologic diagnoses and who were instructed to follow predefined procedures. Cases were summarized and presented as clinical vignettes in randomized order to six raters (three junior and three senior neurologists with particular interest in movement disorders) who were unaware of the clinical design of the study. No specific diagnostic criteria were provided. Clinical vignettes included demographic data, history of present and past illness, therapies, family history, physical and neurologic examinations, and clinical laboratory studies. After reading the information from the first visit, raters completed a standardized form offering their own diagnoses and specifying the main features of the cases. They then evaluated the last visit and suggested a final diagnosis. There were 630 observations (105 cases× 6 raters), including 42 for Pick's.

Statistics. Neurologists' clinical diagnoses were compared with pathologic diagnoses for sensitivity, specificity, and positive predictive values at first and last clinic visits.³⁵ A low prevalence of Pick's in this population was intentionally chosen to mirror closely what neurologists experience in their practices. We chose sensitivity and positive predictive values as outcome measures because the specificity in our study may be artificially high due to the large number of non-Pick's cases included. The Cochran Q test was used to evaluate differences in sensitivity and specificity between raters³⁵ and the McNemar test to evaluate differences in sensitivity and specificity between visits.³⁶ Group reliability was measured with theκ statistics, which take chance into consideration.³⁷ Like a correlation coefficient,κ varies from 0 (chance) to +1.0 (perfect agreement). Strength of agreement is designated poor (κ < 0), slight (0 to 20), fair (0.21 to 0.4), moderate (0.41 to 0.6), substantial (0.61 to 0.8), and near-perfect to perfect (0.81 to 1.0).³⁸ The pooled κ test was used to determine the significance between κ values.³⁷

Logistic regression analysis was performed to identify the variables that could best predict at an early stage the autopsy-confirmed cases of Pick's.³⁹ These variables were all the main variables each neurologist recorded as present when evaluating a clinical vignette. A separate analysis was performed for each of the six raters for the first visit. We calculated the positive and negative predictive values of each rater's model and of the individual main features of the disease.

Results. The reliability for the diagnosis of Pick's for the whole group significantly changed from moderate (κ = 0.42) at the first visit (mean, 53 months after onset) to fair (κ = 0.36, p< 0.05) at the last visit (mean, 78 months after onset). Although agreement was higher among junior raters (κ = 0.6 for the first visit,κ = 0.72 for the last visit) than senior raters (κ = 0.1 for the first visit, κ = 0.2 for the last visit), the differences were not statistically significant.

Sensitivity for Pick's was low (median, 43% for the first visit, 50% for the last visit; range, 0 to 71% for both visits), but specificity was near-perfect (median, 99.4% for the first visit, 100% for the last visit; range, 98.9 to 100% for both visits) (figure). Sensitivity, but not specificity, varied across raters (p < 0.01 for the first visit, p < 0.05 for the last visit). There were no significant differences in sensitivity or specificity between the first and last visits for any rater. The median positive predictive value was 83% for the first visit and 85% for the last visit. Median negative predictive values for both visits were 96% (range, 93 to 97%). The primary neurologists (who followed these patients in life) detected 43% of the cases with Pick's at the first visit and 57% at the last visit.

• Download figure
• Open in new tab
• Download powerpoint

Figure. Validity of the clinical diagnosis of Pick's by six raters (A-F) for the first (A) and last (B) clinic visits.□, individual sensitivity; ○, specificity; [heavy Greek cross], positive predictive value. The average rater sensitivity (40%) was the same at both visits. The average specificity (99.4% at the first visit, 95.4% at the last visit) and positive predictive values (72% at the first visit, 70% at the last visit) changed only minimally from the first visit to the last visit. The mean interval between visits was 25 months.

For both visits, false-negative cases (autopsy-confirmed cases of Pick's confused clinically with other disorders) occurred mainly with AD(table 1). Detailed information on the patients and misdiagnoses at the first visit can be obtained from the National Auxiliary Publications Service. For both visits, false-positive diagnoses (other disorders clinically misdiagnosed as Pick's) were rare and involved only CBD at the first visit and dementia with Lewy bodies at the last visit (seetable 1). The misdiagnosed case of CBD had frontal lobe features without motor abnormalities (i.e., myoclonus, dystonia, asymmetric parkinsonism) at the first visit. The patient was a 67-year-old white woman, presenting with a 3-year history of depression, hostility, progressive"forgetfulness and poor concentration" for 2 years, increased sexual drive for the past year, and poor job performance for 8 months. Desipramine improved mood but not intellectual decline. There was a history of "dementia" in her mother and an aunt. At the first visit, the examination was remarkable for apathy and dis-inhibition. She had normal immediate recall but abnormal short-term and remote memory. The patient exhibited echolalia, anomia and paraphasic errors, frontal lobe-type behavior (decreased fluency, difficulty with abstract thinking, perseveration, impaired judgment, and inappropriate social behavior), and frontal release signs (positive snout, sucking, palmomental, grasping, and pathologic glabellar reflexes). She had agraphestesia but not digital agnosia. The neurologic examination findings were otherwise normal. Three raters confused this case with Pick's.

Of the features recorded at the first visit, logistic regression analysis identified one to three features for all raters, except rater E, that significantly predicted the autopsy-confirmed Pick's patients (e.g., the significant features identified by rater A were early cortical dementia, including severe frontal lobe features) (table 2). Overall, the best predictors included frontal dementia, early cortical dementia (including severe frontal lobe features), absence of apraxia, and absence of gait disturbances at symptom onset. Although the models had good negative predictive values (93 to 96.8%), the positive predictive values of the models were low (0 to 57%). Validity measures for specific clinical features are presented in table 3. Frontal dementia is the most sensitive feature but has both a low specificity and poor predictive value for the diagnosis of Pick's.

Discussion. Our study had several findings. First, neurologists had only fair to moderate interobserver agreement for the diagnosis of Pick's. Second, Pick's was underdiagnosed. Overall, our raters correctly diagnosed only a few cases of Pick's (low sensitivity). Moreover, misdiagnosis of this disorder by primary neurologists who had followed these cases in tertiary centers specializing in the diagnosis of dementia was also low, providing convergent evidence of the underdiagnosis of Pick's in clinical practice. Third, Pick's is often confused with AD. Fourth, false-positive diagnoses were rare and involved CBD at the first visit and dementia with Lewy bodies at the last visit. Finally, frontal dementia, early cortical dementia with severe frontal involvement, absence of apraxia, and absence of gait problems at onset were the best predictors for the diagnosis of Pick's. However, none of the models correctly classified more than 57% of the patients with Pick's. Moreover, most symptoms and signs that are critical for the diagnosis of Pick's appear to be nonspecific. Focusing on the chronology and progression of clinical symptoms and signs, rather than on their presence alone, may be more relevant for a correct diagnosis of Pick's.

Methodologic issues. In our study, the diagnosis of Pick's and non-Pick's was established by comprehensive neuropathologic examination, applying unified histologic diagnostic criteria.¹¹ The small sample of cases of Pick's, which reflects neurologists' experience in actual practice, made it difficult to identify optimal predictors of the diagnosis. Although our study used retrospectively collected data, all raters indicated that the clinical information was complete. Only prospectively collected data will help to minimize the problem of missing data and permit the identification of other clinical features of Pick's that may improve the diagnosis.

Reliability of the diagnosis. Junior neurologists achieved higher agreement for the diagnosis of Pick's (κ = 0.6) than senior neurologists (κ = 0.1). This may be explained by the fact that they are more familiar with general neurology practice than neurologists who specialize in movement disorders. Overall, the raters were in good agreement(κ = 0.56) that patients with autopsy-confirmed Pick's had frontal dementia, but this feature alone was insufficient for a correct diagnosis of the disorder. For example, frontal dementia was recorded as an important feature of some cases, but Pick's was not diagnosed (e.g., rater D). On the other hand, the raters achieved poor agreement when they had to decide whether patients with autopsy-confirmed Pick's had early cortical dementia(memory impairment and aphasia, apraxia, or agnosia; κ = 0.34) or early cortical dementia with severe frontal lobe features (κ = 0.2). Although some of the cases may have been too advanced for this distinction, some raters were unfamiliar with the criteria for frontal dementia (such as the criteria of Lund and Manchester, DSM-IV).^9,40

Because reliability is not a good measure of accuracy (raters may agree but still be wrong), we depended on validity measures that better reflect diagnostic accuracy to examine the disagreements. However, we were somewhat puzzled that the Pick's patients were never correctly classified by all the raters at either the first or the last visits (at least two raters always misdiagnosed a case). This finding supports the premise that misdiagnosis occurs because clinicians are unaware of the characteristic features of Pick's.

It is unclear why interobserver disagreement increased at the last visit. An examination of the validity measures suggests that some raters had more difficulty in diagnosing Pick's at the last visit, thus making consensus less likely. Further assessment is needed to determine whether this finding is related to the biology of the disorders examined (involvement of overlapping anatomic areas and consequently similar neurologic features) or to the analysis of retrospectively collected data that reflect the diagnostic abilities of the primary neurologists.

Validity of the diagnosis. Sensitivity varied among raters; some could detect 60% of the patients, whereas others were unable to detect even one patient (rater D). Overall, the sensitivity of the clinical diagnosis of Pick's was low. The misdiagnosis of this disorder by the primary neurologists suggests that this disorder is underdiagnosed in clinical practice. Underdiagnosis of Pick's is also reflected in a study in which neurologists, at the last visit, detected only 14% of 21 autopsy-confirmed patients.²⁵

On the other hand, awareness of errors in the clinical diagnosis is a step toward improving clinicians' diagnostic skills and research findings. In general, false-negative diagnoses involved mainly AD, followed by Lewy body disease (Parkinson's disease and dementia with Lewy bodies), PSP, and vascular parkinsonism. These results also agree with those of Mendez et al.,^19,25 who found that autopsy-confirmed cases of Pick's were mainly misdiagnosed as AD. Similarly, in another large series of demented elderly subjects whose diagnoses were confirmed by autopsy, two of five cases of Pick's were diagnosed as AD.⁴¹ Thus, misdiagnosis of Pick's impairs the study of patients with AD. Furthermore, the search for the etiopathogenesis and treatment of Pick's is seriously affected.

Why is it so difficult to differentiate these two disorders? Although the frontal dementia syndrome is needed for the diagnosis of Pick's (high sensitivity and negative predictive value), this syndrome is non-specific(low specificity and positive predictive value; see tables 2 and 3). There are many disorders in which a frontal dementia syndrome may be observed, including PSP, AD, dementia with Lewy bodies, or CBD.^42-46 However, in most of these disorders, the frontal dementia syndrome is less severe at the onset or presents at later stages of the disease accompanied by clear manifestations of cortical dementia. If a patient presents with personality changes, such as apathy, euphoria, disinhibition, aberrant motor behavior, or inappropriate social behavior, a diagnosis of Pick's should be seriously considered.⁴⁷ In addition, patients with Pick's tend to have early nonfluent aphasia and echolalia.^19,48 Thus, the progression of symptoms and signs and a combination of features seem more relevant to the diagnosis than the presence of particular features, because most clinical features are nonspecific. In support of this assumption, it is occasionally challenging to differentiate patients with Pick's from those with PSP, because 8% of patients with the latter disorder may also present with cognitive and behavioral disturbances, including florid frontal symptomatology.^42,49 None of our Pick's patients were misdiagnosed as having PSP by all raters at the first and last visits, which suggests that there were always features in the clinical vignette that led some neurologists to make the diagnosis of Pick's. This finding is in contrast to what we observed in PSP or CBD, in which several cases were misdiagnosed as other disorders by all raters at both visits.^31,32 In fact, in the present study, the autopsy-confirmed patient of Pick's that was misdiagnosed as PSP by two of the fix raters had frontal lobe-type features at the onset but did not have ocular motor or postural abnormalities. This patient had other features suggestive of PSP (i.e., symmetric limb rigidity and slowness, axial more than limb rigidity, and frontal lobe-type features), and supranuclear ophthalmoplegia may occasionally be missing in autopsy-confirmed cases of PSP.^49-51 However, four raters recognized that essential features for that diagnosis, such as severe postural instability and supranuclear opthalmoplegia,³¹ were absent. On the other hand, because all our Pick's patients showed striking personality changes at the onset (e.g., disinhibition), neuropsychiatric testing, in addition to neuropsychologic testing, may be necessary for an early correct diagnosis of this disease.⁴⁷ Detailed neuropsychiatric evaluation could differentiate 77% of frontal lobe dementia patients from those with AD.⁴⁷

Although, classically, Pick's is characterized by early changes in personality, deterioration of social skills, and prominent language abnormalities, with initial preservation of memory and praxis,^5,9,40 several Pick's patients in our study exhibited early memory disturbances. Disturbances of memory as an early symptom were recently reported in a large series of Pick's patients who underwent detailed neuropsychological testing.⁵² The series reported by Mendez et al.¹⁹ reflects similar findings, as 60% of their patients had memory disturbances. Whether the memory disturbance exhibited by patients with Pick's is secondary to the frontal or hippocampal lesions observed in this disease needs further clinicopathologic correlation studies.

False-positive diagnosis weakens research studies on biologic markers and specific therapies. Fortunately, false-positive diagnoses were rare in our study. At the first visit, false-positive diagnoses occurred with one case of CBD that was misdiagnosed as Pick's by three of six raters (see Results). Because the spectrum of features of CBD includes frontal lobe dementia as well as primary progressive aphasia, autopsy-confirmed cases of CBD have been confused with Pick's.^53-57 The reverse also occurs; Pick's, particularly the "parietal form," can be misdiagnosed as CBD, because occasionally Pick's presents with a parietal-type symptomatology.^57,58

In our study, only two cases of dementia with Lewy bodies were confused with Pick's at the last visit. Although both patients with dementia with Lewy bodies had frontal lobe dysfunction, they also exhibited features of cortical dementia (i.e., early memory impairment, visual agnosia, ideomotor and dressing apraxia), episodes of acute confusion unrelated to medication, visual hallucinations, and relatively early parkinsonism, features suggestive of dementia with Lewy bodies.³³

Predictor's of Pick's. The predictors identified for the diagnosis of Pick's (i.e., frontal dementia, early cortical dementia with severe frontal lobe features, absence of ideomotor apraxia, and absence of gait problems at onset) add little to what is already recognized as characteristic of Pick's. In fact, the combination of frontal dementia and absence of apraxia are features included in the Lund and Manchester criteria.⁹

The absence of clinical features more specific than frontal lobe dysfunction and the occasional presentation of an advanced stage of the disease at the first visits are obstacles for the diagnosis of Pick's. However, when marked personality changes are present at symptom onset, the diagnosis of Pick's should be considered. Detailed neuropsychiatric evaluation seems promising in distinguishing patients with Pick's from those with AD.^47,59 Although the Kluver-Bucy syndrome has been associated with Pick's,⁶⁰ the classic syndrome may not be frequent in this disease. Whether the systematic inclusion of neuropsychologic, neuropsychiatric, single-photon emission CT, morphometric MRI, or PET^61-65 information would improve the diagnostic accuracy of neurologists needs to be prospectively evaluated.

Note. Readers can obtain 7 pages of supplementary material from the National Academy Publications Service, c/o Microfiche Publications, P.O. Box 3513, Grand Central Station, New York, NY 10163-3513. Request document no. 05379. Remit with your order (not under separate cover), in U.S. funds only, $7.75 for photocopies or $4.00 for microfiche. Outside the United States and Canada, add postage of $4.50 for the first 20 pages and $1.00 for each 10 pages of material thereafter, or $1.75 for the first microfiche and$.50 for each fiche thereafter. There is a $15.00 invoicing charge on all orders filled before payment.

Acknowledgments

We thank Drs. Jean-Jacques Hauw, Susan Daniel, Dennis Dickson, Dikran S. Horoupian, Peter L. Lantos, Kurt Jellinger, and Ann McKee for providing the cases for the NINDS database.