151751-Najiba-Chargi

39 Diagnostics: validation of skeletal muscle mass measurement Table 2. Agreement between estimated and actual low SMM, defined as a LSMI ≤43.2 cm 2 /m 2 Low skeletal muscle mass: actual LSMI ≤43.2 cm 2 /m 2 Sum Yes No Low skeletal muscle mass: estimated LSMI ≤43.2 cm 2 /m 2 Yes 65 (A) 31 (B) 96 PPV=68% No 12 (C) 92 (D) 104 NPV=88% sum 77 123 200 Sens 84% Spec 75% Acc=79% Legend: Sens: sensitivity, Spec: specificity, PPV: positive predictive value, NPV: negative predictive value, Acc: accuracy DISCUSSION There is a need for a robust, easy and widely available SMM quantification tool specifically for HNC patients, to allow for routine assessment of SMM without the need for additional di - agnostics. Swartz et al proposed a measurement of SMA at the level of C3 as an alternative to measurement of SMA at the level of L3, using standard head and neck CT imaging. Our current study shows that measurement of SMA at the level of C3 provides a good estimation of SMA at the level of L3 ( r s = 0.75). Total SMA at the level of C3 had a higher correlation with SMA at the level of L3 than cross-sectional area of paravertebral muscles only ( r s  = 0.75 versus r s = 0.70), which is in agreement with results of a previous study, albeit slightly lower. 14 Using the same multivariate formula as described earlier, in a different set of patients, we found a very good correlation ( r = 0.82) between SMA at the level of C3 and L3. The agreement in identification of patients with low SMMwas moderate and the probability that a patient with low SMM accord- ing to C3 has a low SMM with the L3 method is 68%. A measurement of SMA at the level of C3 provides a good estimation of SMA at the level of L3 and subsequent analysis without the need for additional testing. Interobserver agreement was not further tested in this study; a previous study showed excellent interobserver agreement for SMA measurement at the level of C3. 20 There was some variation in the identification of patients with low SMM based on the estimat- ed LSMI compared to the actual LSMI. The estimated LSMI however was on average -3.9 cm 2 lower than the actual LSMI; classifying more patients as having low SMM than there actually are. Because the cut-off value for low SMM (LSMI ≤43.2 cm 2 /m 2 ) is based on estimated LSMI by use of segmented SMA at the level of C3, other cut-off values for LSMI may apply when seg - mentation of SMA at the level of L3 is performed directly. This may explain the false positive rate of 25.2%. However, we acknowledge that an estimation of SMA at the level of L3 based on SMA at the level of C3 is not ideal and probably is not sufficient in the future as the most accurate estimation of a patient’s total SMM. Indeed, Baracos published an article concluding that using single muscle as a sentinel muscle for whole body SMM is a flawed premise. 21 This problem probably also applies to SMA on a single CT slice as a representation of whole body skeletal muscle volume. 2