Response to ‘Mendelian randomization analysis does not support a role for leptin in multiple sclerosis’

We thank Wu et al. ¹ for commenting on our paper and raising some important issues. It is especially relevant considering their interesting study utilizing Mendelian randomization to investigate the potential causal association between leptin and multiple sclerosis.

First, while it is true that IBM SPSS, in one sense, does not support conditional logistic regression, it is still possible to perform by using the COXREG procedure as described here (https://www.ibm.com/support/pages/conditional-logistic-regression-using-coxreg).

Second, concerning the suggestion by Wu et al. that age and gender should have been incorporated into the logistic model since they are associated with leptin levels, we disagree. It is correct that leptin levels were associated with both variables. However, the cases and controls were matched regarding both age and sex, and it is, therefore, inappropriate to include them in the model when using conditional logistic regression as they do not vary within each case–control set. ²

Third, the last point from Wu and colleagues was that there apparently were no significant associations between genetically predicted leptin levels and multiple sclerosis. The authors base this on gene loci identified in a large genome-wide association study, although these loci likely do not fully explain variations in leptin levels, as 30%–50% of this variation has been reported to be explained by genetic factors. ³ Information on the gene variations included in the analysis and the percentage of measured leptin explained by these loci would be of value. Still, it is an exciting finding which implies that obesity influences the risk of developing multiple sclerosis through pathways other than leptin per se. We cannot rule out the possibility that the associations seen in our study are due to leptin and adiposity being highly correlated. However, it could also be the case that there exists a threshold effect during adolescence that is not easily captured by a Mendelian randomization approach. Genes, although present from birth, may also show an age-dependency in their ability to affect the outcome of interest, either through interactions with environmental factors or through conditions associated with ageing. ⁴ There is also the issue of pleiotropy, that is, the possibility that some of the genetic variations influencing leptin levels have additional effects that modulate multiple sclerosis risk via other pathways. Nevertheless, we fully agree with Wu et al. that more studies are warranted to clarify whether, and if so, how, leptin is involved in multiple sclerosis pathogenesis.