Altered RBC aggregability in diabetes: a threshold for pathophysiological structure-function RBC changes

Abstract

Objective: Vascular complications have been associated with pathophysiological RBC changes including reduced RBC deformability (RBCD): the ability of RBCs to change their shape in the microvasculature and increased red blood cell aggregability (RBCA): the tendency of RBCs to form aggregates (rouleaux). However, understanding of aberrant RBCA in diabetes and its relationship with RBCD is limited. We investigated clinical and RBC structural and functional changes associated with RBCA in comparison with RBCD.

Methods: We conducted an outpatient cross-sectional study of participants with diabetes (n = 81) and nondiabetic controls (n = 78) at the National Institutes of Health. Clinical history was obtained using standardized forms, fasting blood and urine samples were collected for clinical laboratory measurements and ektacytometry studies of RBC physiological parameters: RBCA, RBCD and osmotic fragility. Functional RBC changes were assessed using hemoglobin-oxygen dissociation (p50). Major outcomes assessed differences in RBCA Aggregation Index (AI), with increased RBCA defined as higher AI. Exploratory outcomes assessed differences in subgroups with type 1 (T1D) and type 2 (T2D) diabetes, and associations with clinical, inflammatory, endocrine and RBC physiological parameters. Outcomes were assessed using both nonparametric and multivariable regression methods.

Results: Compared with controls, AI was significantly higher in the diabetes cohort (75±8 vs 71 ± 11, p = 0.007), T2D (77±7 vs 71 ± 11, p < 0.001), but not T1D (69±7 vs 71 ± 11, p = 0.242). While RBCD values were similar in T1D and T2D (p = 0.31), mean AI value was significantly higher in T2D vs T1D (p = 0.02). In contrast to RBCD, AI was associated with T2D (p < 0.001), BMI (p < 0.001), insulin resistance (HOMA-IR, p = 0.033), leptin (p < 0.001), C-peptide levels (p = 0.009), but not vascular complications (p = 0.156). While RBCA and RBCD were associated with CRP (p < 0.001 and p = 0.005 respectively), neither was associated with fibrinogen (p = 0.768 and p = 0.118 respectively). In contrast to RBCD, AI was associated with increased osmotic fragility (p < 0.001) and a threshold of altered hemoglobin-oxygen dissociation (p50) estimated at AI of 75 in both groups using piecewise linear statistical modeling. Thus, AI was positively correlated with p50 at AI < 75 (p = 0.033) and negatively correlated at AI ≥ 75 (p = 0.007).

Conclusions: Findings show that in contrast to RBCD, RBCA is characterized by a RBC functional threshold beyond which oxygen release is reduced, and RBC structure-function relationship likely modulated by insulin resistance and pro-inflammatory endocrine factors. The distinct but complementary and synergistic changes in RBCD and RBCA provide a framework for strategies aimed at mitigation of vascular risk in diabetes.

Supplementary information: The online version contains supplementary material available at 10.1186/s40842-025-00256-2.

Keywords: Deformability; Diabetes; Hemoglobin-oxygen dissociation; Osmotic fragility; RBC aggregability; RBC structure-function; Vascular complications.

Fig. 1

RBCA (AI) differences in diabetes comparator groups, and relationship with RBCD. (A) Mean AI values and estimated shifts in distribution on the Box-Cox transformed scale across comparator groups. Squares indicate Hodges-Lehmann shifts, with horizontal dotted lines indicating 95% confidence interval (CI). The CI margins show either a vertical bar symbol reflecting actual values or an arrow symbol indicating visual margins were abbreviated for visual clarity. For clarity, values are shown in native (untransformed) scale. (B) Change-in-estimate analysis of diabetes differences adjusted for demographic covariates. Squares indicate Hodges-Lehmann shifts, with horizontal dotted lines indicating 95% confidence interval (CI). The CI margins show either a vertical bar symbol reflecting actual values or an arrow symbol indicating visual margins were abbreviated for visual clarity. (C) Relationship between mean RBCA (AI) and RBCD (SS1/2) in diabetes comparator groups and nondiabetic controls. Horizontal lines indicate 95% confidence interval for RBCD (SS1/2), and vertical lines indicate 95% confidence interval for RBCA (AI). Abbreviations: AI, Aggregation Index; DM, diabetes mellitus; nDM, non-diabetic controls

Fig. 2

Associations between RBCA (AI), demographics, clinical, inflammatory and endocrine parameters using diabetes-status-adjusted linear regression on Box-Cox transformed scale for AI in all participants. Association between AI and clinical variables (A), inflammatory and endocrine parameters (B). Squares indicate regression-modeled shifts in mean AI on the Box-Cox transformed scale, with horizontal dotted lines indicating 95% CI. The CI margins show either a vertical bar symbol reflecting actual values or an arrow symbol indicating visual margins were abbreviated for visual clarity. ^a Black and White groups were comprised of non-Hispanic participants. “Other” was comprised of Asians, Hispanic/Latino ethnicity, and multiracial participants. ^b Presence of micro- or macrovascular complications. ^c Type 2 relative to type 1 diabetes. *values were adjusted (divided or multiplied by 10 or 100) to fit the scale of the forest plot. Abbreviations: AI, Aggregation Index; BMI, body mass index; eGFR, estimated glomerular filtration rate; PCR, protein-to-creatinine ratio; HOMA-IR, homeostatic model assessment for insulin resistance

Fig. 3

Relationship between RBCA and hemoglobin-oxygen dissociation (p50) in the diabetes cohort (N = 58) and nondiabetic controls (N = 75). (A) Association between mean RBCA (AI) and mean p50 values in subgroups stratified by ai tertiles within the diabetes cohort (red circles) and nondiabetic controls (blue circles); horizontal dotted lines indicate 95% confidence interval for p50, and vertical lines indicate 95% confidence interval for RBCA (AI). (B) unstratified AI-p50 associations in all participants, back-transformed to the native scales to exhibit statistically significant and qualitatively distinct associations above and below the estimated ai threshold of ≈75. Note that, to meet assumptions inherent to estimating threshold via piecewise regression, we report the distinct positive-then-negative association measures from either side of the threshold as maximum-likelihood estimated slopes linear on the Box-Cox scale (see supplemental methods for more detail). Abbreviations: AI, Aggregation Index; DM, diabetes mellitus; nDM, non-diabetic controls; L: low AI tertile, M: medium AI tertile, H: high AI tertile

Fig. 4

Relationships between RBCA, osmotic fragility (Omin) and RBCD (SS1/2) in the diabetes cohort and nondiabetic controls. Association between mean RBCA (AI), mean osmotic fragility values stratified by (A) and RBCD values (B) in subgroups stratified by AI tertiles within the diabetes cohort (red circles) and nondiabetic controls (blue circles); horizontal dotted lines indicate 95% confidence interval for Omin or SS1/2, and vertical lines indicate 95% confidence interval for RBCA (AI). Abbreviations: AI, Aggregation Index; DM, diabetes mellitus; nDM, non-diabetic controls; RBCA, Red Blood Cell Aggregability; RBCD, Red Blood Cell Deformability

Fig. 5

Scheme illustrating structure-function relationships between RBCA and RBC parameters in diabetes