When less is statistically Equivalent: Lobe-Specific optimization of histopathological assessment in SARS-CoV-2...

Precise histopathological assessment of pulmonary lesions in animal models is fundamental to evaluating COVID-19 interventions. The multifocal, heterogeneous distribution of SARS-CoV-2-induced pathology in rhesus macaques presents a critical challenge: balancing comprehensive evaluation against resource efficiency. No statistically-validated sampling optimization exists for this widely-used model. We hypothesized that lobe-specific, statistically-validated sampling thresholds could maintain assessment accuracy while significantly reducing analytical burden.

We developed a semi-quantitative scoring system targeting interstitial pneumonia-the predominant histopathological feature in SARS-CoV-2-infected rhesus macaques (n = 12). Two ACVP board-certified pathologists independently evaluated 710-1634 high-power fields (40 × magnification) per animal across seven lung lobes, achieving substantial inter-rater reliability (Cohen's κ = 0.74). To determine minimum sampling requirements maintaining statistical equivalence with comprehensive assessment, we employed bootstrapping simulation (10,000 iterations) combined with Two One-Sided Tests (TOST) equivalence analysis (bounds: ±0.25 pathology points).

Optimal sampling percentages exhibited significant lobe-specific variability: left caudal (25 %, p = 0.047), right caudal (30 %, p = 0.038), left/right proximal, (50 %, p = 0.044/p = 0.043), right accessory (50 %, p = 0.172), right middle (60 %, p = 0.049), and left middle (75 %, p = 0.084). Power analysis demonstrated robust detection capability (range: 0.45-0.72) at α = 0.05. These optimized parameters reduce required field assessments by 25-75 % while maintaining statistical equivalence.

This first anatomically-stratified, statistically-validated methodology significantly enhances histopathological assessment efficiency in the rhesus macaque COVID-19 model. By establishing lobe-specific minimum sampling thresholds that preserve statistical equivalence, our approach optimizes resource utilization while maintaining sensitivity to detect intervention effects, potentially accelerates preclinical therapeutic evaluations.