Incidence of Creutzfeldt-Jakob Disease in the United States from 2000 to 2019

Study Design

This is a cross-sectional study using the National Inpatient Sample (NIS), a nationally representative, deidentified survey performed by the Agency for Healthcare Research and Quality. The largest all-payer medical sample in the USA, the NIS includes data on approximately 8 million inpatient hospitalizations each year, representing a 20% stratified sample of all nonfederal US hospitals. Each de-identified discharge record includes one primary diagnosis and up to 14 secondary diagnoses, using the International Classification of Diseases, Ninth and Tenth Editions (ICD-9 and ICD-10). The NIS also provides information regarding patient demographics, comorbidities, procedures, case-severity measures, and hospital characteristics.^12,13 This study was approved by the institutional review board at Weill Cornell Medical College and the need for obtaining informed consent was waived.

Population and Measurements

We included all adult hospitalizations from 2000 to 2019 with a diagnosis of CJD, defined by the ICD-9 discharge diagnosis codes 046.1 (which includes 046.11 and 046.19) or the ICD-10 code A81.0 (which includes A81.00, A81.01, and A81.09) in any diagnosis code position. Prior studies have used the ICD-9 code 046.1^2,8,14–16 and the ICD-10 code A81.0^{8,11,14,16,17} from cause of death or hospital discharge datasets to track CJD diagnosis in several countries.

Statistical Analysis

We used sampling weights provided by the NIS to calculate nationally representative estimates, and population estimates from the US Census to calculate hospitalization rates per 1 million U.S. adults per year.^12,18,19 Because of the NIS sample redesign, we used the updated trend weights for 2000 to 2011 and the original discharge weights for 2012 to 2019. Age-adjusted rates were calculated using 2000 U.S. census as the standard population. We conducted statistical analysis using STATA (version 15.1, College Station, Texas) and Joinpoint Regression Software (version 4.9.1, National Cancer Institute).^18,20 The joinpoint regression model assumes that data may be divided into segments and each segment may have its own trend. Segments are connected at joinpoints. The Joinpoint Regression Software determined the number of significant joinpoints by performing a series of permutation tests and used Monte Carlo methods to calculate a p-value for each test. After the number of joinpoints was determined, the annual percentage change was calculated for each of the detected time segments. The same procedure was used in subgroup analyses to detect sex and race differences in the trend. All trend analyses were 2-sided tests, and statistical significance was defined as p<0.05.

Prior studies of CJD incidence

Our results should be taken in the context of prior work evaluating the incidence of CJD. Several countries have tracked CJD incidence since 1970. A 2020 metanalysis using data predominantly from the CJD International Surveillance Network EuroCJD and the United Kingdom National CJD Research and Surveillance Unit estimated the incidence of CJD to be around 1–2 cases per million per year from 2005 to 2018. The metanalysis noted a steady increase in sporadic CJD cases between 1996 and 2018, with continually low cases of iatrogenic, genetic, and variant forms of CJD.³ The United States Centers for Disease Control and Prevention tracks CJD deaths using a combination of ICD codes with other surveillance mechanisms, and reports an age-adjusted rate of roughly 1 to 1.5 cases per million person-years from 1979 to 2020.²¹ Our study and another recent study, which reported an age-adjusted incidence of 1.4 per million person-years,¹¹ detected an increasing age-adjusted incidence of CJD per million person-years. Other studies have found age-adjusted mortality from CJD ranging from 0.19 to 2.29 during the years 1979 to 2018 in Europe, Australia, Canada, Japan, and the United States,^{2,4–9,14,22–24} with some studies reporting an increase.^6,8,9 In this context, our study provides the novel finding that, although the incidence of CJD did increase for approximately a decade starting in 2004, the incidence appears to have plateaued starting in 2013.

Limitations

Our study has several limitations. First, as we used the NIS to capture cases of CJD, patients who never presented to the hospital were not included. However, as the majority of patients with CJD are hospitalized at some point in time, we think this is unlikely to substantively affect our results.⁸ Second, ICD codes are an imperfect surveillance tool. In Massachusetts from 2000 to 2008, a death certificate diagnosis of CJD, which closely matched hospital discharge dataset diagnosis code 046.1, was 71% sensitive and 75% specific for pathologically confirmed CJD.²⁵ In the United States between 2003 and 2005, 90.7% of suspected prion disease cases for which physicians obtained neuropathologic results were confirmed.¹¹ Similarly, in Australia between 1970 and 1999, the code 046.1 on a death certificate had a sensitivity of 83.0%, and a 11.5% false-positive rate for CJD, yielding a positive predictive value of 88.5%.²⁶ This is likely because histopathological examination, usually following autopsy rather than diagnostic brain biopsy, is the only way to establish a definite diagnosis of sporadic CJD, and not all CJD cases are referred for autopsy.²⁷ Instead, CJD is usually diagnosed using a combination of brain MRI with abnormal cortical or deep nuclei gray matter on the diffusion-weighted imaging (DWI) and apparent diffusion coefficient (ADC) sequences, which is more than 90% sensitive and specific for sporadic CJD, combined with markers for neuronal injury, and, in late stages of the disease, EEG.^1,28 These diagnostic tools are relatively good but imperfect. In a study in Japan, neuropathologic examination during autopsy confirmed 53 or 60 cases of suspected CJD.²⁹ Similarly, in a study of ten countries’ CJD surveillance systems from 1993 to 2006, 73% of all suspected prion disease cases that underwent neuropathological examination were classified as definite cases.⁴ The imperfection of the ICD-9 and ICD-10 codes for CJD is a reflection of the difficulty in diagnosing CJD without biopsy. Lastly, our findings may be affected by changes in hospital coding practices over time, particularly around the transition to ICD-10 in late 2015. However, while this transition may have affected the absolute rates, it seems unlikely to explain the change in trend.

In conclusion, the incidence of CJD increased in the United States from 2000 to 2019, with a significant increase specifically between 2004 and 2013, though the overall case rate remains low.