Question Posted 03/04/14:
Dear ADNI,

Me and my colleagues have carried out a study where we compare the diagnostic ability of CSF Abeta42 and regional AV-45 PET measures. It would be very valuable if someone from ADNI could verify that I have merged the correct data sets.
CSF data: The UPENN 2nd batch analysis of ADNI 1, 2 and GO.
PET data: Regional AV-45 data from the UCBERK data set.
Diagnosis: Diagnostic summary ADNI1, GO, 2 (file name DXSUM_PDXCONV_ADNIALL, variable DXCHANGE).
I am especially worried about the diagnosis variable (76 NL, 186 stable MCI, 98 stable dementia, 4 NL to MCI, 19 MCI to dementia, 4 MCI to NL). It’s very strange that so many patients are stable MCI. Is this really correct?

If I want to merge the data set with demographic variables, which data sets should I use?

Thank you for your support!

Best regards,


Sebastian Palmqvist
Clinical Memory Research Unit
Clinical Sciences, Malmö
Lund University
Response posted 03/04/14 by Susan Landau:
Hi Sebastian,
The question of whether you have the right biomarker datasets for your study depends on a lot of factors that are too complicated to address here. For example, UC Berkeley has an AV45 dataset available, but there is AV45 data available on the ADNI website from other groups as well, so the dataset you choose depends on the goals of your study. I would suggest you take a look at the methods descriptions of the AV45 datasets, which are available right next to the datasets themselves in the Study Data section. The same is true for the CSF data. If you have questions about specific biomarker datasets, you can contact directly the individual group responsible for the data you are interested in. Contact info is available in the methods descriptions for each dataset.

I'm not sure which demographic data you're interested in, but you can see what is available in the Data Dictionary (also available in the Study Data section).

For the diagnostic information, the numbers of subjects for each Dx group (including converting/nonconverting MCIs) is constantly changing since subjects are continuously completing longitudinal assessments. In addition, the number of subjects with CSF and AV45 data available depends on exactly which CSF and AV45 datasets you choose, the time interval between CSF & AV45 measurements (since some subjects do not have concurrent CSF & AV45 measurements), and whether you are including both ADNI1 both ADNIGO/2 subjects.

In general, though, you can find Dx at enrollment in the ARM.csv file. Dx at each subsequent timepoint is available in the DXSUM*.csv file that you mentioned. However, tracking of Dx changed between ADNI1 and ADNI GO/2. For ADNI1, Dx at each timepoint is tracked in the DXCURREN column, and for ADNIGO/2, Dx at each timepoints is tracked in the DXCHANGE column. The codes for each of those columns are defined in the Data Dictionary file.

Hope this helps