Introduction

Creutzfeldt-Jakob disease is a progressive and fatal disorder of the central nervous system; it is a transmissible spongiform encephalopathy.¹ The incidence of Creutzfeldt-Jakob disease is 0.5-1 new case per million population.² Clinically, Creutzfeldt-Jakob disease is characterised by rapidly progressing dementia and more than 90% mortality within one year of onset.¹ Diagnosis can only be made by neuropathological or immunochemical identification of the pathological isoform of the prion protein in human brain tissue.³ The pathological isoform has also been identified in tonsillar tissue at necropsy but this identification has not been used to confirm the diagnosis of Creutzfeldt-Jakob disease before death.⁴ Diagnosis is usually made clinically⁵ and electroencephalographically.⁶ Diagnosis can be confirmed by biochemical analysis of cerebrospinal fluid which identifies neuron specific enolase, ^{7 8} S100 protein, ^{7 9} tau protein,¹⁰ and 14-3-3 protein.^11-13 However, detection of these proteins in serum has not been useful diagnostically for methodological reasons, such as the detection limits for S100 and 14-3-3 protein and problems with cross reactions with non-brain specific isoforms. Thus biological confirmation of the clinical diagnosis while the patient is still alive has been hampered by the inability to easily and repeatedly identify biological markers.

At our clinic we use an immunoluminometric method to detect brain specific S100b protein in serum (LIA-mat Sangtec100, Bromma, Sweden). The detection limit of this assay is 20 pg/ml in serum; earlier assays had detection limits of 0.5 ng/ml. ^{7 9} S100 is an acidic calcium binding protein which is found as a homodimer or heterodimer of two subunits,  and , with molecular weights of 10.4 kD and 10.5 kD, respectively. In brain tissue S100 is found mainly in glial cells. ^{14 15} Raised concentrations of S100 in serum and cerebrospinal fluid have been reported after major¹⁶ and minor head injury.¹⁷ In these cases S100 concentrations decreased rapidly as the protein was eliminated by the kidneys; the estimated biological half life of S100 is 2 hours.¹⁸ S100 was originally thought to be a marker of brain destruction in acute disease, but now it is thought to be a marker of activated astroglia, which is seen at all stages of Creutzfeldt-Jakob disease.³ We investigated the diagnostic potential of measuring serum concentrations of S100 in the differential diagnosis of Creutzfeldt-Jakob disease.

	Subjects and methods

Since June 1993 suspected cases of Creutzfeldt-Jakob disease in Germany have been reported to the Creutzfeldt-Jakob disease surveillance unit at the department of neurology at Georg August University in Göttingen. Each patient who is reported to the unit is visited by a research physician and examined using a standardised protocol¹⁹; a control patient without dementia was examined at the same time. All controls were matched for age and sex with a patient with suspected Creutzfeldt-Jakob disease. Serum samples were obtained from some of the controls.

Between June 1993 and December 1995, 300 patients suspected of having Creutzfeldt-Jakob disease were seen by a member of the research team. Serum samples were obtained from 224 patients; only these patients are included in the present paper. We also analysed serum samples from the first 35 control patients seen.

All suspected cases of Creutzfeldt-Jakob disease were classed as definitely having the disease, probably having the disease, possibly having the disease, or having other disease according to clinical criteria described elsewhere. ^{5 20} All patients had cranial computed tomography or magnetic resonance imaging of the brain, or both, to exclude ischaemic stroke, haemorrhages, or space occupying lesions as a cause of illness.

Patients were classed as probably having Creutzfeldt-Jakob disease if they had rapidly progressive dementia of less than 2 years' duration and periodic sharp wave complexes on electroencephalography. They also had to have any two of the following symptoms: myoclonus; visual or cerebellar symptoms, or both; pyramidal or extrapyramidal signs, or both; or akinetic mutism. Patients who fulfilled the criteria for probably having the disease but did not have any abnormalities on electroencephalography were classed as possibly having the disease. Patients who did not fulfil the criteria for either possibly or probably having the disease were classed as having other disease. Patients who had an identified pathological isoform of the prion protein in brain tissue on immunohistochemical analysis were classed as definitely having the disease.³

Sample collection and analysis
Serum samples were collected by a research physician and stored at80°C within 24 hours.

Statistical analysis
Standard measures of test validity including sensitivity, specificity, and predictive values with 95% confidence intervals were calculated.²² The optimal cut off point for dichotomising serum concentrations of S100 was selected to maximise the Youden index.²³ The receiver operating characteristics curve was calculated to show the variability of sensitivity and specificity for cut off points of different concentrations of S100 protein, which was measured as a continuous variable.²⁴

	Results

Serum concentrations of S100 protein were determined for 224 patients who were suspected of having Creutzfeldt-Jakob disease and for 35 control patients without dementia. Final diagnoses for the 224 patients included 65 patients neuropathologically classed as definitely having the disease, an additional 6 classed as definitely having the disease as a result of a genetic mutation, 43 classed as probably having the disease, 36 as possibly having the disease, and 74 as having other disease. The most recent clinical information including results of neuropathological examination, if available, was used to divide the patients into groups. The distribution of patients by clinical diagnosis and their final classification is shown in table 1. Final diagnosis for patients classed as having other disease was obtained either at necropsy or by follow up reports from hospitals. The distribution of patients and controls by age and sex is shown in table 2.

View larger version (16K): [in this window] [in a new window]   Fig 1.    Serum concentrations of S100 protein by final diagnosis. 10th, 25th, 75th, and 90th percentiles are shown. CJD=Creutzfeldt-Jakob disease; *Patient with meningoencephalitis; Patient with amyotrophic lateral sclerosis

Of the cases of Creutzfeldt-Jakob disease with a genetic basis two had fatal familial insomnia with a mutation at D178N, one had Gerstmann-Sträussler-Scheinker disease with a mutation at P102L, two patients had a mutation at V210I, and one had an insert mutation (9*24). The patients with fatal familial insomnia and the patient with Gerstmann-Sträussler-Scheinker had all been classed clinically as possibly having Creutzfeldt-Jakob disease. Initially, the patients with the mutation at V210I had been classed as probably having the disease; the patient with the insert mutation had initially been classed as having other disease.

For further analysis those classed as definitely and those classed as probably having the disease were grouped together since almost all of those classed as probably having the disease became definite cases after necropsy. Patients' clinical symptoms at the time of entry into the study are shown in table 3.

The serum concentrations of S100 protein in the combined group of patients classed as definitely or probably having the disease ranged from 95 pg/ml to 2016 pg/ml (median 395 pg/ml) (fig 1). Among those who definitely had the disease serum concentrations ranged from 120 pg/ml to 2016 pg/ml (median 439 pg/ml); among those who probably had the disease concentrations ranged from 95 pg/ml to 1199 pg/ml (median 347 pg/ml). For those classed as possibly having the disease concentrations ranged from <20 pg/ml to 742 pg/ml (median 118 pg/ml). Among those classed as having other diseases concentrations ranged from <20 pg/ml to 1198 pg/ml (median 109 pg/ml). Concentrations in the control group ranged from <20 pg/ml to 657 pg/ml (median 97 pg/ml). Additional information on serum concentrations of S100 in those classed as having other diseases is given in table 4 with the final diagnoses. Diagnoses and concentrations are given for patients in the control group without dementia in table 5.

Differences between the three groups (the combined group of those with definite or probable disease, those with possible disease, and those with other disease) were significant (P=0.0001). There was also a significant difference between matched pairs of the combined group (those with definite or probable disease) and the controls (n=35) (P=0.0001). There was no significant difference when the group of those classed as possibly having the disease was compared with those classed as having other disease (P=0.1). There was a significant difference when the combined group (those with definite or probable disease) was compared with those classed as having other disease (P=0.0001). There were no significant differences between those classed as definitely having the disease and those classed as probably having the disease (P=0.1). No differences that were related to the sex of the patient were found.

The best results for sensitivity and specificity were obtained at the cut off point of 213 pg/ml (Youden Index 0.59). The receiver operating characteristics curve is shown in figure 2. The combined group and the group of those classed as having other diseases were used to calculate sensitivity and specificity. Sensitivity at 213 pg/ml was 77.8% (95% confidence interval 68.8% to 85.2%). Specificity at this value was 81.1% (70.3% to 89.3%). The positive predictive value was 85.7% (77.2% to 92%) and the negative predictive value was 71.4% (60.5% to 80.8%). Applying the same cut off point to the control group gave a specificity of 85.9% (69.0% to 94.6%).

View larger version (12K): [in this window] [in a new window]   Fig 2.    Curve of receiver operating characteristics at different cut off points of serum concentration of S100 protein

Fourteen of the 17 patients who were initially classed as possibly having the disease and whose disease was later confirmed neuropathologically had concentrations of S100 protein >213 pg/ml. Another two patients who were initially classed as having other disease and later reclassified as probably or definitely having the disease had concentrations >213 pg/ml. There were significant differences in the Kaplan-Meier survival curves plotted for different serum concentrations of S100 (fig 3) (P=0.023). Patients with fatal familial insomnia had concentrations of <20 pg/ml and 22 pg/ml. The serum concentration of S100 in the patient with an insert mutation was 178 pg/ml. The patient with Gerstmann-Sträussler-Scheinker disease had a serum concentration of 1434 pg/ml. The two patients with mutations at V210I had serum concentrations of 296 pg/ml and 757 pg/ml.

View larger version (16K): [in this window] [in a new window]   Fig 3.    Kaplan-Meier survival curves at different serum concentrations of S100 protein (P=0.023)

	Discussion

We have identified raised concentrations of a biochemical marker, S100 protein, in the serum of patients with Creutzfeldt-Jakob disease. Previously, diagnosis of the disease had been difficult because no biochemical marker had been identified in serum. We have obtained sensitivity of 77.8% and specificity of 81.1% at a cut off point of 213 pg/ml. Serum concentrations of S100 in the group of patients classed as definitely or probably having the disease were significantly different from concentrations in the other patients.

The primary use of this test will be in the differential diagnosis of diseases which cause dementia. Raised serum concentrations of S100 protein are found in acute and subacute neurological diseases such as stroke, subarachnoid haemorrhage, and hypoxia causing brain damage ^{16 17} ; however, these diseases are usually excluded by cerebral computed tomography or magnetic resonance imaging. We found high concentrations of S100 in a patient with multiple sclerosis and in two patients with subacute meningoencephalitis (table 4). These diseases can be differentiated from the diagnosis of Creutzfeldt-Jakob disease by magnetic resonance imaging or lumbar puncture. Raised concentrations of S100 were also found in 14 out of 17 patients initially classed as possibly having Creutzfeldt-Jakob disease but reclassified as definite cases after neuropathological examination. Two out of three patients initially classed as having other disease also had raised concentrations of S100 and were later reclassified as definitely or probably having the disease. Measuring concentrations of S100 protein in serum may aid early diagnosis of Creutzfeldt-Jakob disease, at least until more specific tests are available. Specificity was acceptable when the cut off point of 213 pg/ml was applied to the control group of patients without dementia. However, this test must be evaluated on a larger group of patients who do not have dementia.

The serum test for S100 does not have the same sensitivity and specificity for the diagnosis of Creutzfeldt-Jakob disease as tests on cerebrospinal fluid. ^{8 9 11} In a smaller study of patients with Creutzfeldt-Jakob disease an immunoblot assay for 14-3-3 protein in cerebrospinal fluid had 98% sensitivity and 99% specificity.²⁵ However, other studies using this assay reported more false negatives. ^{13 26} A single measurement of serum concentrations of S100 will not replace a single measurement of proteins in cerebrospinal fluid, but repeated measurement of S100 over intervals of several days may improve diagnostic accuracy. It remains to be seen whether an analysis which combines results from tests on cerebrospinal fluid with results from serum tests would be more useful.

There are significant differences in survival associated with the serum concentration of S100; these differences are shown by the Kaplan-Meier curves. The differences support the hypothesis that serum concentrations of S100 increase during the course of the disease and the quantitative value may provide information about disease progression. Data from serial tests may provide verification of this; there could be alternative explanations for our findings such as that the disease may progress faster in patients with raised concentrations of S100.

A serum test for Creutzfeldt-Jakob disease is preferable to analysis of cerebrospinal fluid. Because of the short biological half life of S100 (2 hours)¹⁸ raised serum concentrations will decrease rapidly in cases of subarachnoid haemorrhage¹⁶ and circulatory arrest during heart surgery,²⁷ but concentrations of S100 will not fall during the course of Creutzfeldt-Jakob disease because of activation of the astroglia. Thus, acute diseases in which serum S100 concentrations rise can easily be excluded from the differential diagnosis by the transience of the rise. Additionally, because of the high homology between human and bovine S100²⁸ the serum test may also be used to diagnose bovine spongiform encephalopathy when there are no clinical signs of the disease.