marieverbanck
commented
2 years ago Hi @may-moore,
Here is a general reply to your comments. MR-PRESSO is a sequential method that comprises three components
1- Global test
MR-PRESSO applies an IVW MR model using all specified IVs and computes a global test for horizontal pleiotropy (H0: there is no horizontal pleiotropy). The global test provides a test statistic (residual sum of squares) and a P-value reported in:
$`MR-PRESSO results`$`Global Test`
2- Outlier test
If horizontal pleiotropy is detected in step 1, MR-PRESSO performs an outlier test for each IV and remove all offending IVs that are due to horizontal pleiotropy before running IVW MR without the detected outliers. The outlier test provides a test statistic (residual sum of squares) and a P-value for each IV reported in:
$`MR-PRESSO results`$`Outlier Test`
3- Distortion test
If horizontal pleiotropy is detected in step 1 and outliers are detected in step 2, MR-PRESSO performs a distortion test to test the difference in the causal estimates before and after outlier removal. The distortion test provides: 1) the list of the 'Outliers Indices' identified as outliers and excluded to calculate 2) the 'Distortion Coefficient' (in percent) and 3) its P-value.
These elements are reported in:
$`MR-PRESSO results`$`Distortion Test`
Synthesis of the results
The estimated causal effect (as well as standard error, test statistic and P-value) are reported for the raw MR analysis and the outlier-corrected MR analysis in
$`Main MR results`
NB: if no horizontal pleiotropy is detected or no outlier IVs are detected MR-PRESSO outputs NAs for the outlier-corrected analysis.
Best, Marie
Hi,
I just had a query about interpretation of the MR-PRESSO results. My understanding of MR-PRESSO was that it runs an IVW MR estimate and pleiotropy detection (Global MR-PRESSO). If the Global MR-PRESSO is significant it will then run outlier correction. If the outlier correction is sufficiently correcting for the pleiotropy then you will get a significant Distortion p value, if not then there is still pleiotropy.
However, when I run MRPRESSO on multiple traits I sometimes get a significant Global MRPRESSO but it does then not run the outlier correction. Can I ask if the above is correct or if I have misunderstood this?
Many thanks, M