Is length an appropriate estimator to characterize pulmonary alveolar capillaries? A critical evaluation in the human lung.

Stereological estimations of total capillary length have been used to characterize changes in the alveolar capillary network (ACN) during developmental processes or pathophysiological conditions. Here, we analyzed whether length estimations are appropriate to describe the 3D nature of the ACN. Semi-thin sections of five human lungs, previously investigated by Gehr et al. (Respir Physiol 1978; 32:121-140), were used to estimate alveolar capillary length using a "design-based" or a "model-based" stereological approach. The design-based approach involves counting of capillary profiles related to a defined area of the reference space. The model-based approach bases on the assumption that capillaries are round tubes and length was calculated from capillary volume and surface area. The model-based approach provided a mean of 6,950 km (SD: 3,108 km) for total capillary length, the design-based approach resulted in a mean of 2,746 km (SD: 722 km). Because of the geometry of the ACN both approaches carry an unpredictable bias. The bias incurred by the design-based approach is proportional to the ratio between radius and length of the capillary segments in the ACN, the number of branching points and the winding of the capillaries. The model-based approach is biased because of the real noncylindrical shape of capillaries and the network structure. In conclusion, the estimation of the total length of capillaries in the ACN cannot be recommended as the geometry of the ACN does not fulfill the requirements for stereological length estimation. Until new methods are being developed, the unbiased estimates of capillary volume, and surface area should be preferred.