An Aging Risk-Factor Scale: Biomarkers of Renal Disease and Anemia are Primary Predictors of Three-Year Survival in Common Marmosets (Callithrix jacchus)

Technical Summary: An Aging Risk-Factor Scale for Common Marmosets

Problem Statement
The common marmoset (Callithrix jacchus) is an increasingly vital nonhuman primate model for geroscience and translational research due to its short lifespan and genetic proximity to humans. However, a critical gap exists in understanding which specific age-related physiological changes prospectively predict survival in captive colonies. While it is established that aging marmosets frequently develop renal insufficiency, anemia, and metabolic declines, it remains unclear which specific biomarker thresholds indicate elevated mortality risk. This uncertainty hinders the design of longitudinal aging studies, the stratification of cohorts for intervention trials, and the implementation of targeted clinical monitoring for colony welfare. The study aims to identify prognostic markers, define empirically derived high-risk thresholds, and develop a composite risk-factor scale to screen for mortality risk in captive marmosets.

Methodology
The study prospectively evaluated 66 adult marmosets (33 males, 33 females; aged 2–16 years) housed at the Southwest National Primate Research Center. Baseline data were collected across four physiological domains:

1. Body Composition and Metabolism: Quantified via EchoMRI (lean mass, fat mass, fluid retention) and flow-through respirometry (Resting Energy Expenditure, REE).
2. Hematology: Complete blood counts (CBC) including erythrogram, leukogram, and platelet parameters.
3. Blood Chemistry: Comprehensive panels including renal function markers (SUN, creatinine), electrolytes, minerals, and liver enzymes.
4. Demographics: Chronological age.

Animals were followed for a 3-year period to determine survival status. Statistical analysis proceeded in three phases:

• Univariable and Multivariable Modeling: Cox proportional hazards models were used to assess continuous predictors. A parsimonious 10-variable multivariable model was selected to avoid multicollinearity and overfitting, representing major physiological domains.
• Threshold Definition: Receiver Operating Characteristic (ROC) curve analysis and Youden's Index were applied to the 10 selected variables to identify optimal cut-points for dichotomizing animals into "low" vs. "high" risk categories.
• Scale Development and Validation: The 10 binary risk factors were summed to create a composite score (0–10). This scale was evaluated using Kaplan–Meier survival analysis, Cox regression, and logistic regression. Additionally, path analysis was employed to map directional relationships among age, body composition, kidney disease, anemia, and metabolic rate.

Key Results

• Survival Outcomes: Of the 66 subjects, 18 did not survive the 3-year period. Necropsy revealed that 83.3% of non-survivors had kidney disease, and 93.3% of those with kidney disease also had anemia.
• Predictive Markers: In univariable analyses, numerous markers predicted survival, including fluid retention, SUN, creatinine, phosphorus, hemoglobin, RBC, and REE. In the fully adjusted multivariable Cox model, only Serum Urea Nitrogen (SUN) and Red Blood Cell Count (RBC) remained independently significant predictors of mortality.
• Risk Thresholds: ROC analysis identified specific high-risk thresholds for the 10 variables, including SUN ≥26 mg/dL, Creatinine ≥0.5 mg/dL, RBC ≤6.69 × 10⁶/µL, and Hemoglobin ≤10.2 g/dL.
• Composite Scale Performance: The 10-item risk scale demonstrated strong predictive power (concordance = 0.835), explaining approximately 42% of the variance in survival. Each additional risk factor increased the hazard of death by 1.75-fold (95% CI: 1.43–2.14).
• Stratification: A threshold of ≥7 risk factors identified a high-risk group with 100% mortality (11/11 deaths) within 3 years and 100% specificity. The scale successfully stratified the population into three groups: low risk (0–1 factors, 0% mortality), medium risk (2–6 factors, 18.9% mortality), and high risk (≥7 factors, 100% mortality).
• Path Analysis: The structural model confirmed that age acts as an upstream driver, predicting lower lean mass and higher SUN. SUN, in turn, was positioned as a central node predicting higher creatinine, greater fluid retention, and lower hemoglobin, supporting a "kidney–anemia–fluid" axis as a primary pathway of vulnerability.

Significance and Claims
The authors present a marmoset-specific screening tool for mortality risk that integrates multi-domain physiological data, noting that the work is currently a preprint. The primary significance lies in the development of a pragmatic, 10-item deficit count scale that functions as a proxy for "biological age" or frailty.

The paper asserts that:

1. Renal and Hematologic Dominance: Renal disease (indicated by SUN) and anemia (indicated by RBC) are the primary drivers of late-life vulnerability in marmosets, forming a convergent pathway that links kidney dysfunction to fluid dysregulation and erythropoietic failure.
2. Clinical Utility: The scale offers a practical tool for colony management. The high specificity of the ≥7 threshold allows for the identification of animals requiring immediate veterinary intervention or exclusion from experimental protocols where imminent death would confound endpoints.
3. Geroscience Application: The scale enables more precise baseline stratification and covariate adjustment for "physiological age" in intervention trials, moving beyond chronological age to assess true biological resilience.
4. Translational Relevance: By mapping the "kidney–anemia–fluid" axis, the findings reinforce the marmoset's utility as a translational model for studying the inter-organ communication and multi-system loss of resilience characteristic of human aging.

The authors maintain a modest tone regarding causality, noting that path analysis suggests plausible structures rather than definitive causal mechanisms. They explicitly state that the ROC-derived thresholds were optimized within the same dataset and require external validation in future cohorts to assess generalizability and recalibrate cut-points if needed.