Temporal trajectories of real-life hospital laboratory data measured during usual care predict clinical outcome in men and women with HFpEF
E Hatip Unlu, T F Kok, M Barry Loncq De Jong, P F Van Den Berg, N Suthahar, R M A Van Der Boon, W Meijers, R A De Boer, J J Brugts, E Boersma, I KardysAbstract
Background
Even though blood biomarkers are linked to clinical outcomes, relying on one or two measurements does not fully capture their dynamic patterns over time. Therefore, we investigated serially measured hospital laboratory values in ambulatory patients with HFpEF to identify prognostic indicators. We addressed the relevance of organ system-specific sets of laboratory values, and explored potential sex differences in these associations.
Methods
We included ambulatory HFpEF patients who attended the outpatient clinic of an academic hospital between 2017 and 2022. Data on 32 laboratory values, several clinical characteristics, and outcomes were extracted from electronic health records. Associations between time-varying laboratory values and all-cause mortality were analysed using joint modelling, including interaction terms to assess potential sex differences.
Results
In 515 patients, median (25th–75th percentile) age was 73 (61, 81) years, and 288 (56%) were women. During a median follow-up of 2.9 (1.6, 4.3) years, a total of 153 (30%) patients were deceased. Several characteristics and laboratory values (age, weight, creatinine, urea, ALT, GGT, ferritin, hemoglobin, leukocytes and thrombocytes), differed between sexes at baseline. Serial measurements of 23 laboratory values, including NT-proBNP, were significantly associated with all-cause mortality (Figure 1). There were no statistically significant sex differences in hazard ratios for NT-proBNP, cardiac troponin T and CRP, while ALT, total bilirubin, HDL cholesterol, LDL cholesterol and triglycerides exhibited statistically significant sex differences. In women only, elevated HDL cholesterol, LDL cholesterol, total protein and iron levels were linked to lower risk; additionally, elevated leukocyte levels were linked to increased risk. Moreover, in men only, elevated triglycerides and thrombocyte levels were linked to a lower risk, while higher ALT and total bilirubin levels were linked to a higher risk (Figure 2). In five different models containing multiple laboratory values, NT-proBNP maintained statistical significance in all models, alongside several others, including albumin, eGFR, AST, GGT, LDH, CRP and hemoglobin. Yet, the incremental discriminative abilities of these models remained limited compared to NT-proBNP only, with the highest AUC reaching 0.724.
Conclusion
This study illustrates the ability of real-life serially measured hospital laboratory data to predict all-cause mortality in men and women in an ambulatory HFpEF population. Sex-based differences were observed in certain serially measured lipid profiles and liver enzymes, and for some parameters, statistical significance persisted exclusively in either men or women. Serially measured NT-proBNP was most robust for risk prediction in these HFpEF patients, and while other laboratory values were also independent predictors, their incremental discriminative value was limited.Figure 1For image description, please refer to the figure legend and surrounding text.Figure 2For image description, please refer to the figure legend and surrounding text.