Challenging the Clinical Utility of the 14-3-3 Protein for the Diagnosis of Sporadic Creutzfeldt-Jakob Disease FREE

THE 14-3-3 protein is a normal neuronal protein that is released into cerebrospinal fluid (CSF) in association with acute neuronal injury.¹ It has been suggested that the presence of 14-3-3 protein in CSF is a reliable marker for Creutzfeldt-Jakob disease (CJD), with sensitivity and specificity for this protein reported as high as 96% and 93% to 100%, respectively.^{2 - 4} These reports have recently led the World Health Organization to revise its diagnostic criteria for probable sporadic CJD (sCJD) to allow substitution of a positive 14-3-3 test for a positive electroencephalogram, provided the disease has less than a 2-year duration.⁵ Yet, a series of studies has suggested that both sensitivity and specificity for sCJD are lacking with this test.^{1 ,6 - 7}

Previous studies that have examined the sensitivity of the 14-3-3 protein for CJD were performed in a single laboratory and did not restrict themselves to pathology-proved cases.^{4 ,8 - 9} In this study, to assess the true sensitivity of the 14-3-3 test, we included only cases that were pathology-proved for sCJD. In addition, because 14-3-3 tests were sent by referring physicians to one or more of several possible laboratories, this study may better reflect typical clinical experience with 14-3-3 in the United States.

We reviewed the medical records of all patients referred to and/or examined at the University of California, San Francisco (UCSF), Departments of Neurology and Pathology (Division of Neuropathology), from January 1, 1998, through July 15, 2002, in whom CJD was pathologically confirmed with either brain biopsy (10 patients), autopsy (18 patients), or both (4 patients). Medical records were reviewed for the results of any 14-3-3 testing. None of the 14-3-3 tests were performed at our institution. We identified 32 patients in whom a biopsy or autopsy confirmed a diagnosis of sCJD and who had had the 14-3-3 test performed. Testing was performed at several laboratories (Mayo Clinic, Rochester, Minn; Case Western Reserve University, Cleveland, Ohio; National Institutes of Health, Bethesda, Md; The Johns Hopkins University, Baltimore, Md; Neuromark Inc, Los Altos, Calif; and Focus Technologies Inc, Cypress, Calif). One 14-3-3 test (patient 8) was performed outside the United States at a European National Surveillance Unit. The 14-3-3 tests were performed by 1 of 2 methods, enzyme-linked immunosorbent assay (ELISA) or immunoblot. Fourteen patients had their 14-3-3 tests performed by ELISA and 18 by standard Western blot using a polyclonal antibody raised against the β isoform of the 14-3-3 protein.^{3 ,10} Three patients (patients 2, 26, and 28) had 14-3-3 tests analyzed by commercial laboratories (Neuromark Inc and Focus Technologies Inc). Patient 2 also had the test sent simultaneously to the National Institutes of Health and Neuromark Inc (with equivalent results). In 19 cases, pathological specimens were reviewed by the UCSF Division of Neuropathology; the remaining 13 cases were analyzed at other institutions and we obtained the pathology report. Eleven of the 32 patients were clinically examined at UCSF and 21 were examined at other institutions.

All 32 cases showed histopathological findings consistent with prion disease, including extensive vacuolation within the neuropil, neuronal loss, and astrogliosis.¹¹ Thirty-one of the 32 patients had some form of positive immunostaining for protease-resistant prion protein (PrP^Sc), by tissue immunoperoxidase staining, histoblot analysis, and/or immunoblot with the monoclonal antibody 3F4 (Santa Cruz Biotechnology Inc, Santa Cruz, Calif), which is specific to the prion protein (PrP).¹¹ One patient (patient 8) did not have PrP immunostaining performed but had histopathological findings and a clinical course consistent with sCJD. Pathology specimens processed at UCSF were immunostained with the 3F4 antibody after denaturing techniques. PrP^Sc was identified by immunoperoxidase staining of formalin-fixed, paraffin-embedded tissue sections by the hydrolytic autoclaving method¹² and/or by immunostaining of unfixed, frozen cryostat sections blotted by nitrocellulose paper—the histoblotting technique.¹¹

Differences between subjects with positive and negative 14-3-3 test results in age, total duration of the disease, the time from first symptom to the lumbar puncture (LP), the time from LP to death, and a ratio of the time from first symptom to LP over total disease duration were carried out with 1-way analysis of variance or χ² analysis with the use of SPSS software (version 10.1.0; SPSS Inc, Chicago, Ill). For the 2 patients who were still alive at the time of our analysis, the disease duration, or survival, was considered up until September 15, 2002. For the 2 patients with initially negative and later positive results, we used the time point of the positive 14-3-3 test result for statistical analyses.

The results of each patient's 14-3-3 test(s) are shown in Table 1. Only 53% (17/32) of the patients diagnosed as having pathologically proved sCJD had a positive 14-3-3 test result. Two patients initially had negative 14-3-3 test results, but on later testing had positive results. The 14-3-3 test results were conservatively considered positive for these 2 patients. One case (patient 9) was initially negative at 4 months, but was positive at 7 months into a 9-month-long illness (performed at the same laboratory). Another case (patient 22) had 2 negative test results (drawn on the same day) at 1.5 months and a positive test result at 2.75 months (all performed at the same laboratory).

One-way analysis of variance (Table 2) indicated a significant difference between the 14-3-3 result and the number of months between disease onset and 14-3-3 testing (LP) (F_1,30 = 5.16, P = .03), with subjects with a positive result being tested much sooner. There were no significant differences between the positive and negative 14-3-3 groups in age, total duration of the disease, the time from the LP test to death, and a ratio of the time from first symptom to the LP over disease duration. In addition, there were no significant differences in sex for patients with negative or positive test results (Pearson χ² = 1.89; P = .17). As shown in Table 3, no single laboratory accounted for the majority of negative 14-3-3 results, and both the ELISA and immunoblot methods contributed to the low sensitivity for the 14-3-3 test. The immunoblot was only 56% sensitive, while the ELISA method was only 43% sensitive. Pathological examination showed no clear differentiating features between the patients who were 14-3-3 positive and those whose test results were negative.

The 14-3-3 testing (by immunoblot or ELISA) had a sensitivity of only 53% in pathology-proved cases of sCJD in this study. We looked for factors that might play a role in determining 14-3-3 results but found no obvious relationship between a positive or negative 14-3-3 result and patients' age, duration of the disease, the time from the 14-3-3 test to death, a ratio of the time from first symptom to the 14-3-3 test over disease duration, the laboratory that performed the 14-3-3 test, or the testing method (immunoblot vs ELISA). There was a significant relationship between a positive 14-3-3 result and a shorter time from disease onset to the test.

Previous evaluation of 14-3-3 protein in patient populations within national CJD surveillance studies has suggested a sensitivity of 93% to 97%^{3 ,13 - 14} and a specificity of 84% to 100%.^{2 - 3 ,13 - 14} When analyses of only pathologically proved cases reported in the literature are considered, the sensitivity of 14-3-3 was lower, ranging from 88% to 100%.^{2 - 3 ,13 - 14} We can only speculate as to why there is a discrepancy between our findings and results of these previous studies. Some possibilities include laboratory error, methodologic differences in the assay, patient selection, or changes in the level of CSF 14-3-3 protein during the course of the illness.⁷ Perhaps testing in various laboratories in the United States is not consistent, even when the same method is used (immunoblot or ELISA). Importantly, there are no universally accepted uniform standards for performing 14-3-3 tests, regardless of method, making the utility of such a broad test worrisome. This is an important consideration, as in some countries all 14-3-3 tests are done in the same laboratory, which is part of a national surveillance center^{8 ,13 - 14} ; this is not the situation in the United States, where 14-3-3 tests are done by government, university, and commercial laboratories. In our analysis, 14-3-3 tests were performed at 7 different locations, all but 1 in the United States. Some investigators have suggested that 14-3-3 tests done at commercial laboratories are not as sensitive as those at academic or government laboratories, but we found no statistical relationship between 14-3-3 results and the laboratory that performed the tests. Only 2 patients' 14-3-3 tests in our analysis were done at a commercial laboratory, so our conclusion that the test is not highly sensitive remains unchanged.

If the 14-3-3 protein is a nonspecific marker for extensive neuronal injury, it is possible that the 14-3-3 test is more likely to be positive later in the disease course, and perhaps with serial testing more patients become positive for the 14-3-3 test. Only 2 patients (patients 9 and 22) were tested at 2 different time points (at both times tests were performed in the same laboratory). Both were initially negative and later became positive. However, our statistical analysis did not show any relationship between when in the course of the illness the test was done (expressed as a ratio of the time from disease onset to the 14-3-3 test over the total disease duration) and the 14-3-3 result. Therefore, it is unlikely that a discrepancy between our results and those of other studies is accounted for by the timing of the test. Whatever the reason, our analysis suggests that, in the United States, the 14-3-3 is not a sensitive test for diagnosing sCJD.

In addition, the 14-3-3 test is not sensitive for other types of prion disease, such as variant and familial CJD.^{13 ,15} Furthermore, the breadth of neurologic diseases reported in the literature that give a positive 14-3-3 result, many of which resemble CJD, indicates that the 14-3-3protein lacks specificity. Many studies and case reports have found positive 14-3-3 test results in nonprion diseases.^{1 ,6 - 8 ,14} Human 14-3-3 proteins are normal neuronal and nonneuronal proteins that play a role in modulation of signal transduction pathways and are released into the CSF when there is nonspecific, rapid, and extensive destruction of brain tissue. Its elevated level in the CSF of patients across a breadth of diseases involving acute neuronal damage suggests that the protein is a nonspecific marker for neuron or astrocyte death.^{1 ,3 ,6} Our finding that patients with a positive 14-3-3 test result tended to have shorter time from disease onset until the 14-3-3 test was performed is consistent with the idea that 14-3-3 is associated with acute neuronal injury. Patients with a more fulminant and/or rapid pace of disease will likely undergo an LP and 14-3-3 test earlier than patients with a slower illness.

If the application of the 14-3-3 test in the diagnosis of sCJD causes physicians to overlook cases because of a false-negative 14-3-3 result, the potential for harm is substantial. Diagnosis of sCJD can help to minimize the risk of iatrogenic transmission from neurosurgical or pathological procedures. In addition, false-negative results prevent patients, caregivers, and physicians from addressing this fatal illness in a realistic fashion.⁷ Efforts to develop minimally invasive prion-specific assays for early diagnosis are currently under way.^{16 - 17} These assays should have better sensitivity and specificity for CJD than the 14-3-3 protein if they are going to be used in clinical practice. We suggest that the 14-3-3 protein does not offer sufficient sensitivity or specificity to be used as a test for CJD.

Corresponding author and reprints: Michael D. Geschwind, MD, PhD, UCSF Memory and Aging Center, Department of Neurology, UCSF Medical Center, 350 Parnassus Ave, Suite 706, Campus Box 1207, San Francisco, CA 94143-1207 (e-mail: mgeschwind@memory.ucsf.edu).

Accepted for publication February 6, 2003.

Author contributions: Study concept and design (Drs Geschwind, D. Miller, and B. L. Miller and Ms Martindale); acquisition of data (Drs Geschwind, D. Miller, DeArmond, Uyehara-Lock, Gaskin, and Barbaro and Ms Martindale); analysis and interpretation of data (Drs Geschwind, D. Miller, DeArmond, Uyehara-Lock, Gaskin, Kramer, Barbaro, and B. L. Miller and Ms Martindale); drafting of the manuscript (Drs Geschwind and B. L. Miller and Ms Martindale); critical revision of the manuscript for important intellectual content (Drs Geschwind, D. Miller, DeArmond, Uyehara-Lock, Gaskin, Kramer, and B. L. Miller and Ms Martindale); statistical expertise (Dr Kramer); administrative, technical, and material support (Ms Martindale and Drs DeArmond, Uyehara-Lock, Gaskin, and Barbaro); study supervision (Drs Geschwind, D. Miller, and B. L. Miller).

This study was supported by the John Douglas French Foundation for Alzheimer's Disease Research, San Francisco; The McBean Family Foundation, San Francisco; and contract N01-NS02328 from the National Institutes of Health, National Institute for Neurological Diseases and Stroke, Bethesda, Md. This study was carried out in part in the General Clinical Research Center, Moffit Hospital, University of California, San Francisco, with funds provided by the National Center for Research Resources, contract M01 RR-00079, US Public Health Service, Washington, DC.

We thank Daniel Geschwind, MD, PhD, and Howard Rosen, MD, for their critical review of the manuscript.