Quantitation of free light chains in the cerebrospinal fluid reliably predicts their intrathecal synthesis

Abstract

Background

The results of free light chains quantitation in the cerebrospinal fluid were recently compared with the presence of cerebrospinal fluid-restricted oligoclonal IgG, but not oligoclonal free kappa light chains and oligoclonal free lambda light chains. We therefore aimed to compare the performance of the quantitative tests with the qualitative one for the same molecule.

Methods

Seventy-five paired cerebrospinal fluid and serum samples were analysed for oligoclonal IgG, oligoclonal free kappa light chains and oligoclonal free lambda light chains. Cerebrospinal fluid and serum free kappa and lambda light chains were quantified using Freelite™ kits on SPA Plus analyzer. ROC curves were analysed for the prediction of intrathecal synthesis and compared for cerebrospinal fluid concentration, cerebrospinal fluid/serum quotient (Q_fLC) and index (Q_fLC/Q_Albumin). The presence of cerebrospinal fluid-restricted oligoclonal free kappa light chains and oligoclonal free lambda light chains bands was used as reference.

Results

No statistically significant differences were observed among cerebrospinal fluid concentration, Q_fLC and index for the prediction of free light chain intrathecal synthesis. Each parameter was able to predict the occurrence of cerebrospinal fluid-restricted oligoclonal free light chain bands (AUCs 0.932–0.999). However, we noted elevated cerebrospinal fluid free light chain concentrations in the absence of cerebrospinal fluid-restricted oligoclonal free light chain bands in two patients with very high serum free light chain values.

Conclusions

Quantitation of cerebrospinal fluid free light chains reliably predicts their intrathecal synthesis. Yet, cerebrospinal fluid/serum quotient may still be preferred to correct for high serum free light chain concentrations. An appropriate formula should be sought to correct for blood–cerebrospinal fluid barrier status.

Introduction

Several recent reports propose cerebrospinal fluid (CSF) free light chain (fLC) measurement as a valuable marker of intrathecal immunoglobulin synthesis that has predictive value similar to oligoclonal IgG (o-IgG).^1–4 Up to now, CSF fLC measurements have only been compared to o-IgG as a gold standard. This approach, however, does not compare exactly like with like. We therefore aimed to compare the performance of the quantitative tests with the qualitative one for the same molecule, i.e. oligoclonal free kappa light chains (fKLC) and oligoclonal free lambda light chains (fLLC).

Methods

Seventy-five paired CSF and serum samples under o-IgG analysis were also analysed for o-fKLC and o-fLLC, as described elsewhere. ⁵ Samples were then kept frozen at −30℃ until quantitative fLC measurements. CSF and serum fKLC and fLLC concentrations were determined on the SPA Plus turbidimetric analyser using dedicated Freelite™ kits (The Binding Site Ltd.) CE marked for CSF analysis (Ref. No. LK016.L.S and LK018.L.S.). Statistical analysis was performed using MedCalc Software. Presence of CSF-restricted o-fLC and o-IgG bands was used as a reference. CSF samples measured as <0.1 mg/L were analysed as 0.09 mg/L. P values < 0.05 were considered significant.

Results

Using cut-offs for CSF-restricted bands (≥4 for IgG, ≥6 for fKLC and ≥2 for fLCC) determined in our laboratory, ⁵ 26 samples were positive for o-IgG, 30 for o-fKLC and 28 for o-fLLC. All samples positive for o-IgG were positive for o-fKLC; three samples positive for o-IgG were negative for o-fLLC and five samples negative for o-IgG were positive for o-fLLC.

CSF fKLC concentrations varied between <0.1 and 1.33 mg/L (median: 0.20; interquartile range: 0.11–0.27) in the o-fKLC negative group and between 0.80 and 16.90 mg/L (6.62; 2.71–11.61) in the o-fKLC positive group. CSF fLLC concentrations varied between <0.1 and 1.44 mg/L (0.22; 0.17–0.29) in the o-fLLC negative group and 0.24 and 12.03 mg/L (1.55; 0.56–3.08) in the o-fLLC positive group. Among patients without CSF-restricted o-fLC bands, one patient had high CSF (1.33 and 1.06 mg/L) as well as serum (97.63 and 84.48 mg/L) fKLC and fLLC concentrations and one patient with detectable fLLC paraproteinaemia had high CSF (1.44 mg/L) and serum (223.05 mg/L) fLLC values (see Supplementary File 1).

Analysis of ROC curves for the prediction of appropriate o-fLC positivity showed no significant differences between CSF fLC concentration only or CSF/Serum Quotient (Q_fLC) or index (Q_fLC/Q_Albumin). AUCs for fKLC and fLLC were 0.998 (95% confidence interval: 0.948–1.000) and 0.954 (0.879–0.989), respectively, for CSF concentrations; 0.999 (0.949–1.000) and 0.977 (0.913–0.998) for Q_fLC; 0.991 (0.935–1.000) and 0.932 (0.848–0.997) for fLC index. Associated criteria were as follows: CSF fKLC > 0.56 mg/L; CSF fLLC > 0.43 mg/L;Q_fKLC > 39.2 10⁻³; Q_fLLC> 36.3 10⁻³; fKLC index >6.07; fLLC index >6.68.

When ROC curves for CSF fLC quantitative data as predictors of intrathecal IgG synthesis were compared, CSF fKLC performed best: AUC = 0.997 (95% confidence interval: 0.946–1.000; criterion: >1.38 mg/L), followed by Q_fKLC: AUC = 0.994 (0.940–1.000; >70.7·10⁻³), fKLC index: AUC = 0.989 (0.931–1.000; >7.67) and fLLC index: AUC = 0.978 (0.914–0.998; >5.58). AUCs for CSF fKLC, Q_fKLC, fKLC index and fLLC index proved significantly larger than those for CSF fLLC and Q_fLLC; other differences were not significant (see Supplementary File 2).

CSF but not serum fKLC/fLLC ratio was significantly higher in samples positive for o-IgG (median, 3.137; interquartile range, 1.929–16.482) as compared to negative ones (0.938; 0.679–1.157) (Mann-Whitney test, P < 0.0001).

Although fKLC measurements were able to predict intrathecal IgG synthesis, such a prediction is not perfectly appropriate since o-fLC are sometimes detected in the absence of o-IgG and o-fKLC are more frequently observed in inflammatory CNS diseases than o-fLLC. ⁵ Nevertheless, we proved that CSF fLC quantitation can predict intrathecal fLC synthesis. Although we did not aim at diagnosis-related analysis, it is of note that two patients with CNS inflammation proved positive for both o-fLC, while negative for o-IgG; 19/20 patients with MS or CIS (95%) were both o-IgG and o-fKLC positive and 17 (85%) were also o-fLLC positive. There was almost perfect agreement between qualitative and quantitative fLC data in this subgroup (see Supplementary File 1).

fLC index did not perform better than Q_fLC or CSF fLC only. This led us to investigate the relation of Q_fLC to Q_Albumin in more detail (Figure 1).

Figure 1.

Quotient diagrams for fKLC (A) and fLLC (B) with local regression smoothing trend lines.

Proposing Reiber’s hyperbolic function for the estimation of intrathecal fLC synthesis^1,2 is a misconception. ⁶ Since fLCs have a smaller molecule than albumin, theoretically an inverse function should be used i.e.

$Q_{\mathrm{fLC}}=\sqrt{[\frac{b^2}{a^2}.(Q_{\mathrm{Alb}}+c^2)-b^2]}=b\sqrt{(\frac{Q_{\mathrm{Alb}}+c}{a}{)}^2-1}$

Larger data set, however, is needed to verify our hypothesis (see Supplementary File 3).

Conclusions

Quantitation of fLCs in CSF enables reliable prediction of intrathecal fLC synthesis. Q_fLC may be preferred to correct for elevated serum fLC concentrations. Correction for blood–CSF barrier status seems reasonable, but the most suitable formula remains to be determined. We further suggest that both fLC be analysed together because of substantial differences in the CSF fKLC/fLLC ratios among o-fLC positive patients that might be of some (although yet unknown) significance.