Quantifying Experimental Variability in Shear-Induced Hemolysis to Support Uncertainty-Aware Hemolysis Models

Abstract

Purpose: Numerical hemolysis models rely on experimental data to fit parameters and predict hemolysis under various conditions. However, existing experiments often use few replicates per condition, leaving inherent variability largely unaddressed. This can lead to oversimplified models that fail to capture the true nature of hemolysis. Here, we quantify intra- and inter-donor variability at a single, well-defined shear stress and exposure time and examine how sample size affects measurement precision METHODS: Human blood from five healthy donors was subjected to a fixed shear stress and exposure time condition. For each donor, 20 independent measurements were performed to calculate a hemolysis index (HI). Intra-donor variability (variation within a single donor's measurements) and inter-donor variability (variation between donor means) were compared. Additionally, bootstrap analyses were used to explore the effect of the sample size on the confidence intervals of the mean HI.

Results: Intra-donor variability was approximately four times higher than inter-donor variability, indicating that most of the uncertainty originated from within a single donor's set of samples rather than between donors. Increasing the sample size from 2 to 20 replicates substantially narrowed the confidence intervals of the mean hemolysis estimate, suggesting that commonly used small sample sizes may underrepresent the true variability in hemolysis measurements.

Conclusion: Intra-donor variability is a significant driver of uncertainty in hemolysis measurements at a fixed shear stress and exposure time condition, surpassing differences among donors. Obtaining robust and reliable hemolysis estimates requires increasing the number of replicate measurements to reduce uncertainty. Integrating these insights into future experimental designs and uncertainty-aware hemolysis models will improve the reliability of in silico predictions and inform safer, more effective blood-contacting medical device designs.

Keywords: Experimental hemolysis; Inter-variability; Intra-variability; Uncertainty quantification.

Fig. 1

Schematic representation of the Mooney-Ewart shearing geometry (A), featuring a rotating inner bob and a static outer cylinder (gray). The blood sample volume is illustrated in red. Nominal dimensions of the geometry are provided, with the actual measured dimensions after manufacturing shown in parentheses. The acceleration and deceleration profile of the measurement system is depicted in (B), illustrating the rotational speed (ω) ramping up to 3000 rpm. The period of constant rotational speed, lasting 15 seconds, is defined as the exposure time

Fig. 2

Chronologically ordered hemolysis index (HI) measurements for each donor are shown as bar charts in panels A–E, with 20 individual gray bars numbered 1–20 for each donor. Panel F displays the combined histogram of all HI measurements across all donors, represented in orange

Fig. 3

Hemolysis index (HI) measurements in percentage for each donor, displayed as box plots. The lower and upper edges of each box represent the 25th and 75th percentiles, respectively, while the whiskers indicate the 95th percentile. The mean value is shown as a horizontal line within the box and is also provided numerically along with the standard deviation in the description of the abscissa for each donor. Individual measurements are represented by scatter points overlaid on the box plots

Fig. 4

Bootstrapped distributions of possible mean values for a theoretical sample size of 3 for each donor are shown in panels A–E. The distributions are represented by gray histograms and approximated using a Gaussian kernel density estimate (orange). The black dashed line indicates the mean value for each donor based on the original 20 samples measured in this study. Panel F displays the bootstrapped confidence interval progression across sample sizes ranging from 2 to 20 for all combined samples (orange), with the total mean of all donors represented as a gray line