Diagnosis of Creutzfeldt-Jakob disease by measurement of S100 protein in serum

Editor—Otto et al describe significant differences in the serum concentrations of S100 protein in patients with and without Creutzfeldt-Jakob disease.¹ The population of patients studied is, however, not appropriate for assessing the usefulness of measuring serum concentrations of S100 protein as a diagnostic test, and Otto et al’s calculations of positive and negative predictive values of the test are invalid as they fail to take into consideration the prevalence of Creutzfeldt-Jakob disease as a cause of dementia.

Patients with clinically evident disease do not need a diagnostic test. It is not informative to evaluate the potential of a blood test to distinguish those with clinically definite Creutzfeldt-Jakob disease from those without the disease on clinical grounds. The test is useful only if it helps to predict the likelihood of the disease in a population that does not meet the criteria for probable Creutzfeldt-Jakob disease, which is therefore the population in which it must be assessed.

Otto et al say that the primary use of serum concentrations of S100 protein will be in the differential diagnosis of diseases that cause dementia. At a cut off of 213 pg/ml they calculate a sensitivity of 77.8% and specificity of 81.1% for detecting Creutzfeldt-Jakob disease in a referred population of 108 patients with definite or probable Creutzfeldt-Jakob disease and 74 patients with dementia without Creutzfeldt-Jakob disease (table). When calculating the positive and negative predictive values of a test the prevalence of the target disorder within the population to be examined must be considered.² There are no criteria for distinguishing the subgroup of patients with dementia but without Creutzfeldt-Jakob disease (who in Otto et al’s study were referred to the surveillance unit for Creutzfeldt-Jakob disease) from other patients with dementia. We must therefore consider them to represent all patients with dementia. If the relative prevalence of Creutzfeldt-Jakob disease in dementia is taken to be 1 per 1000 people aged 60-70—a considerable overestimation—the number of false positive results in the population with dementia increases enormously (table). The positive predictive value of the test falls from 85.7% to 0.4%. The negative predictive value rises from 71.4% to 99.97%. This test therefore has no use in this population.

Finally, we agree with Pocchiari that measurement of serum cencentrations of S100 protein may have a role in the diagnosis of Creutzfeldt-Jakob disease.³ Further studies are required, however.

Table.

Sensitivity, specificity, and predictive values of serum concentrations of S100 protein in Creutzfeldt-Jakob disease (CJD) among patients with dementia. Predictive values are dramatically influenced by prevalence of target disorder within the population tested

	Otto et al’s study*		Expected false positives and true negatives for S100 test if prevalence of CJD in dementia were 0.1%† (n=100 800−108)
S100 protein (pg/ml)	CJD present (n=108)	CJD absent (n=74)
>213	84	14	20 412
<213	24	60	87 480
Sensitivity	84/108=77.8%
Specificity		60/74=81.1%
Total	108	74	107 892

^*Prevalence of CJD (108/182)=59.3%. Positive predictive value=true positives/(true+false positives)=84/(84+14)=85.7%. Negative predictive value=true negatives/(true+false negatives)=60/(60+24)=71.4%.  

^†Prevalence of CJD in dementia of 0.1%. Positive predictive value=true positives/(true+false positives)=84/(84+20 412)=0.4%. Negative predictive value=true negatives/(true+false negatives)=87 480/(87 480+24)=99.97%..