We performed a series of two-sample summary level univariable (UV) and multivariable (MV) Mendelian randomization (MR) analyses to investigate whether adiposity-associated proteins mediate the relationship between adiposity and colorectal cancer (CRC).
All scripts were run sequentially and are numbered accordingly. The CRC data was not publicly available and so all data/results which contain data from these summary statistics are not available here (e.g., outcome data files are not made available for the association between adiposity and CRC).
We used three measures of adiposity: body mass index (BMI), waist hip ratio (WHR), and WHR adjusted for BMI (WHRadjBMI). Summary statistics were available from Pulit et al., (2020). Sex-combined and sex-specific summary statistics were used for all three measures.
In the main analysis, summary statistics for 4,907 proteins (Somascan) were available from Ferkingstad et al., (2019) - only sex-combined data were available. In the additional analysis, summary statistics for 1,463 proteins (Olink) were available from Sun et al., (2022) - only sex-combined data were available.
Summary statistics for 16 colorectal cancer outcomes were available from Huyghe et al., (2019), this included: overall, overall measured in Europeans only, overall male, overall female, overall colon, overall proximal, overall distal, overall rectal, female colon, male colon, female proximal, male proximal, female distal, male distal, female rectal, and male rectal. In addition, summary statistics for early onset colorectal cancer were available from XXX et al., (). These data are nto publicly available and can instead be requested from the study authors.
This study has four main analyses that were performed sequentially to estimate: (I) the association between adiposity measures and CRC, (II) the association between adiposity measures and proteins, (III) the association between adiposity-associated proteins (identified in step II) and CRC, and (IV) the potential mediating effect of adiposity-associated proteins in the adiposity CRC association (identified in step II and III). The reverse association was tested for each step. Where proteins were used as the exposure, cis and trans variants were used in the main analysis. For all proteins included in step IV, the following additional analyses were performed where data were available: (I) cis-only MR of the protein-CRC pair, (II) colocalization of the protein-CRC pair, and (III) UVMR using cis and trans variants from a second protein dataset.