Integrative Analysis of Drug Co-Prescriptions in Peritoneal Dialysis Reveals Molecular Targets and Novel Strategies for Intervention

Abstract

Background/Objectives: Peritoneal dialysis (PD) is a renal replacement therapy for patients with kidney failure. Managing PD patients often involves addressing a complex interplay of comorbidities and complications, necessitating the use of multiple medications. This study aimed to systematically characterize commonly co-prescribed drugs in PD and to identify novel drug combinations that may target dysregulated molecular mechanisms associated with PD's pathophysiology. Methods: We analyzed clinical records from 702 PD patients spanning 30 years, encompassing over 5500 prescription points. Using network-based modeling techniques, we assessed drug co-prescription patterns, clinical outcomes, and longitudinal treatment trends. To explore potential drug repurposing opportunities, we constructed a molecular network model of PD based on a consolidated transcriptomics dataset and integrated this with drug-target interaction information. Results: We found commonly prescribed drugs such as furosemide, sucroferric oxyhydroxide, calcitriol, darbepoetin alfa, and aluminum hydroxide to be integral components of PD patient management, prescribed in over 30% of PD patients. The molecular-network-based approach found combinations of drugs like theophylline, fluoxetine, celecoxib, and amitriptyline to possibly have synergistic effects and to target dysregulated molecules of PD-related pathomechanisms. Two further distinct categories of drugs emerged as particularly interesting in our study: selective serotonin reuptake inhibitors (SSRIs), which were found to modulate molecules implicated in peritoneal fibrosis, and vascular endothelial growth factor (VEGF) inhibitors, which exhibit anti-fibrotic properties that are potentially useful for PD. Conclusions: This comprehensive exploration of drug co-prescriptions in the context of PD-related pathomechanisms provides valuable insights for opening future therapeutic strategies and identifying new targets for drug repurposing.

Keywords: biological networks; drug combination; drug repurposing; network analysis; peritoneal dialysis.

Figure 1

Patient flowchart of inclusion for analysis.

Figure 2

Drugs prescribed to individual PD patients over three decades: (A) Length of the lines: time range for which individual patients were treated with PD; color: mean unique drugs prescribed during the treatment period. (B) Over the decades of the investigated timeframe, the overall number of prescribed drugs per patient increased (black dashed line), and also when analyzed per endpoint type: solid violet: kidney transplantation; solid red: transfer to hemodialysis (HD); solid turquoise: death of the patient. (C) Black dots: individual patients; left: patients with the endpoint kidney transplantation (KTx) were treated with PD for longer compared to the other two endpoints, while they were prescribed a lower number of drugs (** p < 0.001, Wilcoxon signed-rank test, adjusted). (D) Number of prescribed drugs is correlated with the prescription points per patient; each dot represents an individual patient.

Figure 3

Three groups of different drug co-prescription patterns, based on the frequency of co-prescription and individual prescription to patients: (A) Color indicates prescription to >30% of patients: red: both drugs; green: one of the two drugs; blue: neither of the drugs. (B) Further detail on the group with both drugs > 30% (red dots in (A)). The fractions of the circles depict the ratio of the percentages of the two drugs. (C) Heatmap of the most prescribed drug combinations. Left side: individual drug prescriptions; triangles: percentage of prescriptions for each individual drug. Right side: percentage of patients co-prescribed a specific drug combination.

Figure 4

Construction of the PD model: (A) In a previous study, we developed a comprehensive PD model using a curated set of dysregulated molecules associated with the most pertinent complications (blue). From these molecules, we elucidated the key dysregulated biological processes (orange), with a particular focus on those relevant to inflammation and angiogenesis. (B) To enhance the model’s precision, we incorporated genes associated with the clinically most important pathophysiological processes, inflammation and angiogenesis, sourced from the Gene Ontology (GO) database (green). Drug targets based on information from the ChEMBL database are shown as pink triangles. (C) Additionally, we pinpointed specific drug targets that simultaneously influence both inflammation and angiogenesis biological processes. GO-derived information is in green, ChEMBL-derived information is in pink, and biological processes are in brown.

Figure 5

Drugs closely aligned with the PD disease network: All drugs concurrently targeting biological processes related to angiogenesis and inflammation through shared or direct targets. Each color corresponds to drugs sharing identical direct drug targets (table bottom left with targeted genes); color-coded lines indicate the structural similarities between paired drugs; frames: already prescribed in our PD patient cohort.