Adrenal venous sampling in primary aldosteronism: multinomial regression modeling to detect aldosterone secretion...

Difficulty to recognize or canulate the right adrenal vein is the most frequent cause of adrenal venous sampling (AVS) failure. We aimed to assess multinomial regression modeling (MRM) of peripheral and left adrenal vein samplings to detect lateralization of aldosterone secretion when the right AVS is missing.

Simultaneous bilateral AVS samplings were performed before (basal) and after intravenous cosyntropin injection in 188 consecutive patients between December 1989 and September 2015. Different reference standards for lateralization of aldosterone secretion were defined for basal and for postcosyntropin AVS and according to lateralization index cutoffs at least 2 and at least 4. MRMs were built to detect lateralization of aldosterone secretion according to these reference standards using only peripheral and left adrenal veins samplings (without the right AVS). Detection accuracy was assessed by the area under the receiver operating characteristic (AUROC) curves and detection sensitivities were reported for specificity at least 95%.

For basal AVS with lateralization index at least 2, AUROC were respectively 0.931 [95% confidence interval (CI) 0.894-0.968] and 0.922 (95% CI 0.882-0.962) for right and left lateralization of aldosterone secretion detection and MRM could detect respectively 65.5 and 62.7% of the right and left lateralization of aldosterone secretion. For AVS after cosyntropin with lateralization index at least 4, AUROC were respectively 0.964 (95% CI: 0.940-0.987) and 0.955 (95% CI: 0.927-0.983) for right and left lateralization of aldosterone secretion, and MRM could detect respectively 77.2 and 72.9% of the right and left lateralization of aldosterone secretion.

MRM can detect lateralization of aldosterone secretion without the right AVS in most patients and could eliminate the need for repeat AVS when right adrenal vein canulation is nonselective or impossible.