Abstract
Results of neuroimaging datasets aggregated from multiple sites may be biased by site-specific profiles in participants’ demographic and clinical characteristics, as well as MRI acquisition protocols and scanning platforms. We compared the impact of four different harmonization methods on results obtained from analyses of cortical thickness data: (1) linear mixed-effects model (LME) that models site-specific random intercepts (LMEINT), (2) LME that models both site-specific random intercepts and age-related random slopes (LMEINT+SLP), (3) ComBat, and (4) ComBat with a generalized additive model (ComBat-GAM). Our test case for comparing harmonization methods was cortical thickness data aggregated from 29 sites, which included 1,340 cases with posttraumatic stress disorder (PTSD) (6.2–81.8 years old) and 2,057 trauma-exposed controls without PTSD (6.3–85.2 years old). We found that, compared to the other data harmonization methods, data processed with ComBat-GAM was more sensitive to the detection of significant case-control differences (Χ2(3) = 63.704, p < 0.001) as well as case-control differences in age-related cortical thinning (Χ2(3) = 12.082, p = 0.007). Both ComBat and ComBat-GAM outperformed LME methods in detecting sex differences (Χ2(3) = 9.114, p = 0.028) in regional cortical thickness. ComBat-GAM also led to stronger estimates of age-related declines in cortical thickness (corrected p-values < 0.001), stronger estimates of case-related cortical thickness reduction (corrected p-values < 0.001), weaker estimates of age-related declines in cortical thickness in cases than controls (corrected p-values < 0.001), stronger estimates of cortical thickness reduction in females than males (corrected p-values < 0.001), and stronger estimates of cortical thickness reduction in females relative to males in cases than controls (corrected p-values < 0.001). Our results support the use of ComBat-GAM to minimize confounds and increase statistical power when harmonizing data with non-linear effects, and the use of either ComBat or ComBat-GAM for harmonizing data with linear effects.
Original language | English |
---|---|
Article number | 119509 |
Journal | NEUROIMAGE |
Volume | 261 |
DOIs | |
Publication status | Published - 1 Nov 2022 |
Keywords
- ComBat
- ComBat-GAM
- Cortical Thickness
- Data Harmonization
- General Additive Model
- Linear Mixed-Effects Model
- PTSD
- Scanner Effects
- Site Effects
Access to Document
Other files and links
Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver
}
In: NEUROIMAGE, Vol. 261, 119509, 01.11.2022.
Research output: Contribution to journal › Article › Academic › peer-review
TY - JOUR
T1 - A comparison of methods to harmonize cortical thickness measurements across scanners and sites
AU - Sun, Delin
AU - Rakesh, Gopalkumar
AU - Haswell, Courtney C.
AU - Logue, Mark
AU - Baird, C. Lexi
AU - O'Leary, Erin N.
AU - Cotton, Andrew S.
AU - Xie, Hong
AU - Tamburrino, Marijo
AU - Chen, Tian
AU - Dennis, Emily L.
AU - Jahanshad, Neda
AU - Salminen, Lauren E.
AU - Thomopoulos, Sophia I.
AU - Rashid, Faisal
AU - Ching, Christopher R. K.
AU - Koch, Saskia B. J.
AU - Frijling, Jessie L.
AU - Nawijn, Laura
AU - van Zuiden, Mirjam
AU - Zhu, Xi
AU - Suarez-Jimenez, Benjamin
AU - Sierk, Anika
AU - Walter, Henrik
AU - Manthey, Antje
AU - Stevens, Jennifer S.
AU - Fani, Negar
AU - van Rooij, Sanne J. H.
AU - Stein, Murray
AU - Bomyea, Jessica
AU - Koerte, Inga K.
AU - Choi, Kyle
AU - van der Werff, Steven J. A.
AU - Vermeiren, Robert R. J. M.
AU - Herzog, Julia
AU - Lebois, Lauren A. M.
AU - Baker, Justin T.
AU - Olson, Elizabeth A.
AU - Straube, Thomas
AU - Korgaonkar, Mayuresh S.
AU - Andrew, Elpiniki
AU - Zhu, Ye
AU - Li, Gen
AU - Ipser, Jonathan
AU - Hudson, Anna R.
AU - Peverill, Matthew
AU - Sambrook, Kelly
AU - Gordon, Evan
AU - Baugh, Lee
AU - Forster, Gina
AU - Simons, Raluca M.
AU - Simons, Jeffrey S.
AU - Magnotta, Vincent
AU - Maron-Katz, Adi
AU - du Plessis, Stefan
AU - Disner, Seth G.
AU - Davenport, Nicholas
AU - Grupe, Daniel W.
AU - Nitschke, Jack B.
AU - deRoon-Cassini, Terri A.
AU - Fitzgerald, Jacklynn M.
AU - Krystal, John H.
AU - Levy, Ifat
AU - Olff, Miranda
AU - Veltman, Dick J.
AU - Wang, Li
AU - Neria, Yuval
AU - de Bellis, Michael D.
AU - Jovanovic, Tanja
AU - Daniels, Judith K.
AU - Shenton, Martha
AU - van de Wee, Nic J. A.
AU - Schmahl, Christian
AU - Kaufman, Milissa L.
AU - Rosso, Isabelle M.
AU - Sponheim, Scott R.
AU - Hofmann, David Bernd
AU - Bryant, Richard A.
AU - Fercho, Kelene A.
AU - Stein, Dan J.
AU - Mueller, Sven C.
AU - Hosseini, Bobak
AU - Phan, K. Luan
AU - McLaughlin, Katie A.
AU - Davidson, Richard J.
AU - Larson, Christine L.
AU - May, Geoffrey
AU - Nelson, Steven M.
AU - Abdallah, Chadi G.
AU - Gomaa, Hassaan
AU - Etkin, Amit
AU - Seedat, Soraya
AU - Harpaz-Rotem, Ilan
AU - Liberzon, Israel
AU - van Erp, Theo G. M.
AU - Quidé, Yann
AU - Wang, Xin
AU - Thompson, Paul M.
AU - Morey, Rajendra A.
N1 - Funding Information: DoD W81XWH-10-1-0925; Center for Brain and Behavior Research Pilot Grant; South Dakota Governor's Research Center Grant; CX001600 VA CDA ; NHMRC Program Grant # 1073041 ; R01 MH111671 ; VISN6 MIRECC ; German Research Foundation grant to J. K. Daniels ( DA 1222/4-1 and WA 1539/8-2 ); VA RR&D 1IK2RX000709 ; NIMH R01-MH043454 ; NIMH K01-MH122774 ; NIMH K01 MH118428-01 (Suarez-Jimenez); 5U01AA021681-08; K24MH71434; K24 DA028773; R01 MH63407; K99NS096116; VA RR&D 1K1RX002325; VA RR&D 1K2RX002922; MH101380; ZonMw, the Netherlands organization for Health Research and Development grant to Miranda Olff ( 40-00812-98-10041 ); Academic Medical Center Research Council grant to Miranda Olff ( 110614 ); VA CSR&D 1IK2CX001680; VISN17 Center of Excellence pilot funding; NIMH R01MH105535 ; NIMH 1R21MH102634 ; German Federal Ministry of Education and Research (BMBF RELEASE 01KR1303A ); German Research Society (Deutsche Forschungsgemeinschaft, DFG; SFB/TRR 58: C06, C07 ); R01MH117601 ; R01AG059874 ; MJFF 14848 ; MH098212 ; MH071537 ; M01RR00039 ; UL1TR000454 ; HD071982 ; HD085850 ; R21MH112956 ; Anonymous Women's Health Fund; Kasparian Fund; Trauma Scholars Fund; Barlow Family Fund; NIMH K01 MH118467 ; W81XWH-08-2-0159 ; Department of Veterans Affairs via support for the National Center for PTSD; NIAAA via its support for (P50) Center for the Translational Neuroscience of Alcohol; NCATS via its support of (CTSA) Yale Center for Clinical Investigation ; NIH R01 MH106574 ; F32MH109274 ; NIMH 1R21MH102634 ; R01MH113574 ; R01-MH103291 ; BOF 2–4 year project to Sven C. Mueller (01J05415); R01MH105355 ; Dana Foundation (to Dr. Nitschke); the University of Wisconsin Institute for Clinical and Translational Research ; a National Science Foundation Graduate Research Fellowship (to Dr. Grupe); the National Institute of Mental Health (NIMH) R01 MH63407 (to De Bellis), R01 AA12479 (to De Bellis), and R01 MH61744 (to De Bellis); R01-MH043454 and T32-MH018931 (to Dr. Davidson); core grant to the Waisman Center from the National Institute of Child Health and Human Development ( P30-HD003352 ); NIMH K23MH112873 ; Veterans Affairs Merit Review Program (10/01/08 – 09/30/13); L30 MH114379; South African Medical Research Council “SHARED ROOTS” Flagship Project; Grant -RFA-FSP-01-2013/SHARED ROOTS; South African Research Chair in PTSD from the Department of Science and Technology and the National Research Foundation ; US Department of Defense Grant W81XWH08-2-0159 (PI: Stein, Murray B); VA RR&D I01RX000622 ; CDMRP W81XWH-08-2-0038 ; South African Medical Research Council ; NARSAD Young Investigator ; Department of Defense award number W81XWH-12-2-0012 ; ENIGMA was also supported in part by U54 EB020403 from the Big Data to Knowledge (BD2K) program; R56AG058854; R01MH116147;; P41 EB015922; 1R01MH110483; 1R21 MH098198; R01MH105355-01A. The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs, the United States Government, or any other funding sources listed here. Publisher Copyright: © 2022
PY - 2022/11/1
Y1 - 2022/11/1
N2 - Results of neuroimaging datasets aggregated from multiple sites may be biased by site-specific profiles in participants’ demographic and clinical characteristics, as well as MRI acquisition protocols and scanning platforms. We compared the impact of four different harmonization methods on results obtained from analyses of cortical thickness data: (1) linear mixed-effects model (LME) that models site-specific random intercepts (LMEINT), (2) LME that models both site-specific random intercepts and age-related random slopes (LMEINT+SLP), (3) ComBat, and (4) ComBat with a generalized additive model (ComBat-GAM). Our test case for comparing harmonization methods was cortical thickness data aggregated from 29 sites, which included 1,340 cases with posttraumatic stress disorder (PTSD) (6.2–81.8 years old) and 2,057 trauma-exposed controls without PTSD (6.3–85.2 years old). We found that, compared to the other data harmonization methods, data processed with ComBat-GAM was more sensitive to the detection of significant case-control differences (Χ2(3) = 63.704, p < 0.001) as well as case-control differences in age-related cortical thinning (Χ2(3) = 12.082, p = 0.007). Both ComBat and ComBat-GAM outperformed LME methods in detecting sex differences (Χ2(3) = 9.114, p = 0.028) in regional cortical thickness. ComBat-GAM also led to stronger estimates of age-related declines in cortical thickness (corrected p-values < 0.001), stronger estimates of case-related cortical thickness reduction (corrected p-values < 0.001), weaker estimates of age-related declines in cortical thickness in cases than controls (corrected p-values < 0.001), stronger estimates of cortical thickness reduction in females than males (corrected p-values < 0.001), and stronger estimates of cortical thickness reduction in females relative to males in cases than controls (corrected p-values < 0.001). Our results support the use of ComBat-GAM to minimize confounds and increase statistical power when harmonizing data with non-linear effects, and the use of either ComBat or ComBat-GAM for harmonizing data with linear effects.
AB - Results of neuroimaging datasets aggregated from multiple sites may be biased by site-specific profiles in participants’ demographic and clinical characteristics, as well as MRI acquisition protocols and scanning platforms. We compared the impact of four different harmonization methods on results obtained from analyses of cortical thickness data: (1) linear mixed-effects model (LME) that models site-specific random intercepts (LMEINT), (2) LME that models both site-specific random intercepts and age-related random slopes (LMEINT+SLP), (3) ComBat, and (4) ComBat with a generalized additive model (ComBat-GAM). Our test case for comparing harmonization methods was cortical thickness data aggregated from 29 sites, which included 1,340 cases with posttraumatic stress disorder (PTSD) (6.2–81.8 years old) and 2,057 trauma-exposed controls without PTSD (6.3–85.2 years old). We found that, compared to the other data harmonization methods, data processed with ComBat-GAM was more sensitive to the detection of significant case-control differences (Χ2(3) = 63.704, p < 0.001) as well as case-control differences in age-related cortical thinning (Χ2(3) = 12.082, p = 0.007). Both ComBat and ComBat-GAM outperformed LME methods in detecting sex differences (Χ2(3) = 9.114, p = 0.028) in regional cortical thickness. ComBat-GAM also led to stronger estimates of age-related declines in cortical thickness (corrected p-values < 0.001), stronger estimates of case-related cortical thickness reduction (corrected p-values < 0.001), weaker estimates of age-related declines in cortical thickness in cases than controls (corrected p-values < 0.001), stronger estimates of cortical thickness reduction in females than males (corrected p-values < 0.001), and stronger estimates of cortical thickness reduction in females relative to males in cases than controls (corrected p-values < 0.001). Our results support the use of ComBat-GAM to minimize confounds and increase statistical power when harmonizing data with non-linear effects, and the use of either ComBat or ComBat-GAM for harmonizing data with linear effects.
KW - ComBat
KW - ComBat-GAM
KW - Cortical Thickness
KW - Data Harmonization
KW - General Additive Model
KW - Linear Mixed-Effects Model
KW - PTSD
KW - Scanner Effects
KW - Site Effects
UR - http://www.scopus.com/inward/record.url?scp=85135967385&partnerID=8YFLogxK
U2 - https://doi.org/10.1016/j.neuroimage.2022.119509
DO - https://doi.org/10.1016/j.neuroimage.2022.119509
M3 - Article
C2 - 35917919
SN - 1053-8119
VL - 261
JO - NEUROIMAGE
JF - NEUROIMAGE
M1 - 119509
ER -