Evaluating Diagnostic Performance of Light Chain Dimerization By Mass Spectrometry to Distinguish between AL Amyloidosis and Monoclonal Gammopathy of Undetermined Significance

Introduction:

Light-chain (AL) amyloidosis often presents with nonspecific symptoms and definitive diagnosis requires a biopsy, which is invasive and often is inconclusive and can result in patients being diagnosed at more advanced stages of the disease. Prior literature investigating free light chain monomer-dimer pattern via western blot demonstrated a high sensitivity and negative predictive value for AL amyloidosis. In this study, we evaluated the utility of light-chain dimer (LCD) levels via a novel mass spectrometry assay (Dimerite) to differentiate between AL amyloidosis and Monoclonal Gammopathy of Undetermined Significance (MGUS).

Methods:

We identified patients with biopsy-confirmed AL amyloidosis and MGUS from samples within our institutional bank. To minimize confounding variables, patients were matched by age, gender, race and involved light chain. Electronic medical records were reviewed for clinical variables, including standard-of-care serum free light chain (sFLC) testing. Samples were analyzed in a blinded fashion by Rapid Novor (Kitchener, ON) for light chain dimerization. Patient serum was analyzed on an Orbitrap Exploris mass spectrometer after digestion by Lys-C and separation of disulfide-bridged peptides by liquid chromatography. LCD values reported as a ratio to control serum. A linear model was developed to define the relationship between sFLC and LCD, with a logarithmic transformation of variables performed due to skewness. Logistic regression was used to assess the discriminant value of LCD at classifying AL vs. MGUS.

Results:

Thirty pairs of AL amyloid and MGUS patients were included in the analysis. The mean age was 66, 63% were male, and 83% were White. Lambda was the involved sFLC in 60% of the cohort. Among AL patients, the heart was the most commonly involved organ (67%), followed by kidney (53%), gastrointestinal tract (20%), and liver (10%). Median involved sFLC was 3.6 mg/dL (SD 30.9) in MGUS compared to 24.7 mg/dL in AL amyloid (SD 81.0, p=0.02). Median uninvolved sFLC was 1.8 mg/dL (SD 5.0) in MGUS compared to 1.7 mg/dL in AL amyloid (SD 4.5, p=0.62). Median involved LCD (expressed as a ratio to control) was 2.4 in MGUS compared to 3.1 in AL amyloid (p=0.13). Median uninvolved LCD was 1.4 in MGUS compared to 0.98 in AL amyloid (p=0.30). LCD and sFLC were strongly correlated (r=0.64) in both involved (r = 0.58) and uninvolved (r = 0.50) subsets (p<0.001).

Given the correlation between sFLC and LCD, we created a linear model to characterize the relationship between involved sFLC values and LCD levels, as well as evaluate if other key variables (light chain (LC) isotype and diagnosis) impacted this relationship. This linear model found that lambda isotype (β=0.59, p<0.001) and involved sFLC values (β=0.73, p<0.001) were associated with higher involved LCD, while diagnosis (AL vs. MGUS) was not (β=-0.09, p=0.39).

We also evaluated several models to predict diagnosis (AL vs. MGUS) using sFLC (involved and uninvolved), LCD (involved and uninvolved) and isotype. An initial model including involved sFLC, uninvolved sFLC and isotype was 76.7% accurate in differentiating AL amyloid and MGUS. The addition of involved LCD did not improve model accuracy, while uninvolved LCD marginally improved model performance (78.3%; p = 0.012). Interestingly, the uninvolved LCD was borderline associated with AL amyloidosis (adjusted odds ratio [aOR] 0.41; p=0.052) while involved sFLC had a much weaker association (aOR = 0.88; p = 0.20), suggesting that dimerization of polyclonal LCs may be disrupted in more advanced disease states. Due to the strong correlation between sFLC and LCD, we created an additional model excluding uninvolved sFLC. The model that included involved sFLC, involved LCD, uninvolved LCD, and isotype was 78.3% accurate and each 1-unit decrease in uninvolved LCD was associated with a 185% increase in odds of having a diagnosis of AL (aOR 2.85; p = 0.026).

Conclusions:

In our cohort, we found that sFLC and LCD were strongly correlated, and LCD level was not associated with the diagnosis of AL amyloid after adjusting for other variables. However, the availability of data on LCD levels was associated with a small but statistically significant improvement in classification between AL and MGUS. The utility of this assay for diagnosis in other plasma cell dyscrasias and longitudinal disease monitoring in AL amyloidosis remain areas of future research.