Abstract
Diabetes mellitus (DM) has been associated with an increased risk of Parkinson’s disease (PD), but most studies have important methodological limitations that may bias the magnitude of this association. DM has also been linked to greater symptom severity and faster disease progression in people with PD, although evidence is largely limited to small clinical or research cohorts with short follow-up, providing little insight into long-term, real-world outcomes. Shared mechanisms between DM and PD have prompted interest in the potential neuroprotective effects of glucose-lowering medications. Although numerous population-based studies have examined medication use in relation to PD risk, there is limited real-world evidence on their effects among people already living with PD.
This thesis examines the association between DM, glucose-lowering medications, PD risk, and long-term outcomes among people with PD using nationwide administrative data from Aotearoa New Zealand, with particular attention to key methodological limitations in the existing literature. Four population-based cohort studies were conducted using Statistics New Zealand’s Integrated Data Infrastructure (IDI) with up to 14 years of follow-up. The first study assessed the association between DM and PD risk in a cohort of people with incident DM, using exposure density sampling and marginal structural models to account for changes in DM status and covariates over time. The second study examined the association between insulin use and PD risk in a cohort aligned by treatment stage, comparing individuals with DM who initiated either insulin or a non-insulin class after two previous non-insulin glucose-lowering classes. The third study assessed the impact of DM on long-term outcomes, including falls, dementia, admission to a long-term care facility (LTCF), and mortality, in individuals with PD with and without DM, and the general population groups without PD with and without DM. The fourth study assessed whether use of metformin, pioglitazone, and sulfonylureas influenced long-term outcomes among people living with both PD and DM.
Findings from the first study showed that DM was associated with a modest increase in PD risk (HR 1.18, 95% CI 1.13, 1.23), which was substantially attenuated after accounting for time-dependent confounding (HR 1.07, 95% CI 1.02, 1.12). In the second study, people with DM who initiated insulin had a lower risk of PD than those who initiated a non-insulin glucose-lowering class (HR 0.71, 95% CI 0.55, 0.93). In the third study, both DM and PD were independently associated with all outcomes, with effects that were sub-multiplicative for most outcomes. Individuals with both PD and DM still experienced reduced survival (HR 1.16, 95% CI 1.09, 1.23), greater risk of recurrent falls (HR 1.13, 95% CI 1.01, 1.26), and earlier admission to an LTCF (HR 1.14, 95% CI 1.06, 1.22), compared to those with PD without DM, while dementia risk did not differ. The fourth study found that, among individuals with both DM and PD, metformin use was associated with lower mortality (HR 0.87, 95% CI 0.79, 0.95) and reduced risk of experiencing recurrent falls (HR 0.83, 95% CI 0.70, 0.99), while pioglitazone was also associated with lower mortality (HR 0.70, 95% CI 0.49, 1.00) and delayed admission to an LTCF (HR 0.52, 95% CI 0.27, 0.98).
This thesis illustrates the value of linked administrative data for examining long-term outcomes and real-world treatment effects in PD populations. These findings highlight the clinical importance of recognising and managing metabolic health in people with PD and support further investigation of metabolic pathways and glucose-lowering medications as potential targets for PD prevention and treatment.