Towards subgroup-specific risk estimates: A meta-analysis of longitudinal studies on olfactory dysfunction and risk of Parkinson's disease.

Abstract Background Interest has risen in identifying individuals at high risk of incident Parkinson’s disease (PD) to facilitate inclusion in neuroprotective treatment trials. Current risk estimates of prodromal markers are based on aggregated data of an entire population, but this approach disregards differences in risk estimates by subgroups of a population. In this proof of concept, we determine subgroup-specific risk estimates of olfactory dysfunction for incident PD. Methods PubMed, EMBASE and Cochrane were searched for prospective studies investigating the association between olfactory dysfunction and incident PD. Random-effects meta-analysis, subgroup analyses and meta-regression were performed to investigate general and subgroup risk estimates. Results Individuals with odor identification dysfunction seemed to be at greater risk of incident PD compared to controls without olfactory dysfunction (OR=4.18; 95%CI [2.47-7.07]). Risk estimates were higher in studies that included higher percentages of women (regression slope s=0.053 increase in log odds ratio per unit increase, p=0.0006), increased with mean study age (s=0.21;p=0.005) and in REM-sleep behavior disorder cohorts (s=1.95;p=0.03). Furthermore, the association between olfactory dysfunction and incident PD was most distinct in studies with shorter follow-up duration (s=-0.56;p=0.0047). Conclusion The presence of olfactory dysfunction conveys a considerably elevated risk of incident PD, likely more in studies with a higher proportion of women, older individuals or short follow-up duration. Individual patient data are warranted to confirm these findings and to yield subgroup-specific risk estimates of other common markers to refine prodromal PD criteria.