Joan A. Casey, PhD, MA1,2; Holly Elser, MD, PhD3,4
To the Editor On behalf of our coauthors, we are writing to report an error in our Original Investigation, “Wildfire Smoke Exposure and Incident Dementia,” published online on November 25, 2024, and in the January 2025 issue of JAMA Neurology.1 We conducted the study to evaluate the association of long-term exposure to particulate matter less than 2.5 microns in diameter (PM2.5) from wildfire and nonwildfire sources with incident dementia diagnoses using an open cohort design at Kaiser Permanente Southern California (KPSC) from 2009 to 2019.
In preparing for a follow-up study for which we extended the cohort through 2023, we detected a coding error that affected our dementia outcome definition. Specifically, we identified dementia using International Classification of Diseases, Ninth Revision (ICD-9) and Tenth Revision (ICD-10) codes for Alzheimer disease, frontotemporal dementia, Lewy body dementia, nonspecific dementia, vascular dementia, and Parkinson disease dementia. To satisfy criteria for Parkinson disease dementia, we required a Parkinson disease diagnosis with a secondary dementia diagnosis at the same encounter (ie, ICD-9: 332.0 and 294.10 or 294.11; ICD-10: G20 and F02.80 or F02.81). When updating our analyses, we found that we incorrectly included individuals with a Parkinson disease diagnosis or a secondary dementia diagnosis (eg, ICD-9: 332.0, 294.10, or 294.11) and therefore inadvertently included individuals with Parkinson disease but with no recorded dementia diagnosis as dementia cases in our previous analysis. This error resulted in the inclusion of 5413 individuals without dementia diagnoses as dementia cases. Because we excluded individuals with prevalent dementia diagnosis at baseline, we also incorrectly excluded from the cohort an additional 5345 KPSC members with a prevalent diagnosis of Parkinson disease.
We have recalculated all analyses using the corrected cohort of 1 228 452 KPSC members (originally reported as 1 223 107), including 75 471 members who developed dementia (originally reported as 80 884). Our previously reported finding of an association between 1-μg/m3 higher 3-year mean wildfire PM2.5 exposure and incident dementia is no longer statistically significant, with a smaller effect size. The adjusted odds ratio (OR) is now 1.12 (95% CI, 0.98-1.28) compared to the previously reported OR of 1.18 (95% CI, 1.03-1.34). The main effect for a 1-μg/m3 higher 3-year mean nonwildfire PM2.5 exposure is similar in terms of effect size and statistical precision (original analysis: OR, 1.01; 95% CI, 1.01-1.02; corrected analysis: OR, 1.007; 95% CI, 1.002-1.011). When evaluating an interquartile range higher 3-year mean wildfire and nonwildfire PM2.5 exposure, we also observed marginally attenuated results (original wildfire PM2.5: OR, 1.02; 95% CI, 1.00-1.03; corrected wildfire PM2.5: OR, 1.01; 95% CI, 1.00-1.03; original nonwildfire PM2.5: OR, 1.03; 95% CI, 1.02-1.04; corrected nonwildfire PM2.5: OR, 1.02; 95% CI, 1.01-1.03).
We also observed changes in models assessing nonlinear exposure-response associations. In the corrected analysis, we observe a linear wildfire PM2.5–dementia association (original: nearly linear). For nonwildfire PM2.5, results remained stable (original: increased up to approximately 6.5 μg/m3, flattened through approximately 13 μg/m3, and then increased), except for a flattening of the association after 17.5 μg/m3.
Except for age and the “Other” racial and ethnic subgroup (the subset of members who reported their race and ethnicity as multiple races, Native American and Alaskan Native, Pacific Islander, other, and unknown race and ethnicity), findings from subgroup analyses are also different from the originally published results. Two wildfire PM2.5 subgroup associations no longer meet the conventional threshold for statistical significance (P < .05) in the corrected analysis. These include the association between 3-year mean wildfire PM2.5 exposure and incident dementia in men (original: OR, 1.28; 95% CI, 1.05-1.56; corrected: OR, 1.21; 95% CI, 0.98-1.48) and high-poverty communities (original: OR, 1.30; 95% CI, 1.04-1.62; corrected: OR, 1.22; 95% CI, 0.97-1.53). Further, we previously observed associations between 3-year mean nonwildfire PM2.5 and incident dementia in both sexes and in high- and low-poverty neighborhoods, but now only find associations in men (OR, 1.01; 95% CI, 1.01-1.02).
Finally, we observed some minor changes in associations exploring alternative exposure metrics for wildfire PM2.5. The association remained unchanged for 1 additional smoke wave (≥2 consecutive days with mean daily wildfire PM2.5 concentration >15 μg/m3; OR, 1.03; 95% CI, 1.01-1.05) and 10-μg/m3 higher wildfire PM2.5 concentration during the peak exposure week (OR, 1.02; 95% CI, 1.00-1.05). The association with an interquartile range higher number of days with wildfire PM2.5 greater than 0 was attenuated and still not significant (original: OR, 1.01; 95% CI, 0.99-1.03; corrected: OR, 1.00; 95% CI, 0.98-1.03), and the association for 1 additional week where mean wildfire PM2.5 exceeded 5 μg/m3 became slightly stronger (original: OR, 1.01; 95% CI, 0.99-1.02; corrected: OR, 1.01; 95% CI, 1.00-1.02).
Because our original findings are no longer statistically significant for the previously reported association between exposure PM2.5 from wildfire sources and incident dementia diagnoses, we have requested that our article be retracted and replaced with a corrected version.
We apologize to the readers of JAMA Neurology for this error and any confusion it has caused. We have had 4 authors independently review all updated results. The corrections affect the cohort composition, cumulative outcome incidence, ORs, and confidence intervals in the Abstract, the Results section, the Table, and Figure 2. In addition, eTables 2-5 and eFigures 1-3 in Supplement 1 were affected. PDF copies of the original article with errors highlighted and with corrections highlighted are now included in Supplement 3 and Supplement 4, respectively.
Corresponding Author: Joan A. Casey, PhD, MA, University of Washington School of Public Health, 3980 15th Ave NE, Seattle, WA 98105 (jacasey@uw.edu).
Published Online: June 30, 2025. doi:10.1001/jamaneurol.2025.2148
Conflict of Interest Disclosures: Dr Casey reported grants from the National Institute on Aging (R01AG071024) during the conduct of the study. No other disclosures were reported.
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.