Abstract by Niklas Andersson

Thiazide diuretics are widely used to treat high blood pressure and are among the first-line antihypertensive treatment options in many countries. Hyponatremia, defined as plasma sodium level <135 mmol/L, is a known but poorly characterized adverse drug reaction to thiazide treatment. This includes uncertainties about the actual frequency and whether certain patients have a particularly high risk, potentially determined by their phenotypic and/or genotypic characteristics.

The array of Danish nationwide demography and healthcare registers available for research makes Denmark a unique setting for population-level observational research and serves as the data foundation for the three papers presented in this thesis. Data hallmarks include their longevity and granularity, the routine and mandatory manner in which data are collected within the Danish free-of-charge social and healthcare system, and the ability to readily link with other data sources, such as biobanks, at the individual level.

In paper I, we emulated pragmatic target trials to estimate the increase in the two-year cumulative incidence of hyponatremia with initiating use of thiazide diuretics compared to non-thiazide antihypertensive drugs (moderate-to-severe hyponatremia, <130 mmol/L, was the primary outcome). We find an excess risk for hyponatremia with thiazide use that is substantially higher than reported in thiazide drug labels and that the risk is particularly pronounced during the first months of treatment and for older patients or those with comorbidities.

In paper II, we explored the heterogeneity in individualized excess risk of moderate-to-severe hyponatremia with thiazide use, aiming to develop and validate causal forest prediction models that leverage phenotypic baseline information to identify high-excess-risk patient groups. We find that a parsimonious causal forest model can efficiently identify high-excess-risk patient groups based solely on information about age and baseline plasma sodium, hemoglobin, and C-reactive protein levels. By targeting a relatively small group of patients with high excess risk to an alternative non-thiazide antihypertensive treatment, we estimate that the burden of thiazide-induced hyponatremia can be markedly reduced at the population level.

In paper III, we examined the genetic architecture contributing to sodium and water homeostasis and for a potential genetic predisposition to thiazide-induced hyponatremia. We map a catalog of loci robustly associated with baseline plasma sodium levels and underline the central role of the pituitary gland in regulating osmolality. We find no evidence of genetic variants specifically predisposing patients to thiazide-induced hyponatremia nor any potential added clinical benefit from preemptive genetic testing to enhance the individual-level risk predictions achieved using baseline phenotypic information alone.