Data analytics improves the diagnostic accuracy of serum free light chain results for detecting monoclonal gammopathy

Abstract

Objectives

To evaluate the real-world performance and reference intervals of the Binding Site Freelite serum free light chain (SFLC) assay (Thermo Fisher Scientific), a global standard for diagnosis, prognostication, and response assessment for monoclonal gammopathies.

Methods

An informatics-based approach was used to retrospectively evaluate concordance between SFLC and the orthogonal Sebia HYDRASYS immunofixation assay results in a large clinical data set consecutively reported between 2010 and 2020.

Results

Among patients with monoclonal-negative results by both SFLC and Sebia HYDRASYS immunofixation assays, 25% (1226/5057) had κ/λ ratios (KLRs) outside the manufacturer-defined and International Myeloma Working Group–cited normal reference interval of 0.26 to 1.65. These results were consistent over the study period and were not affected by sex, age, impaired kidney function, or assay antisera lot variation. Assay drift, in addition to other potential factors, affected the KLR distribution. Using International Statistical Classification of Diseases (ICD) codes, kidney function data, and the central 95% of KLR values generated on the Optilite platform (Thermo Fisher Scientific), we derived a new reference interval of 0.67 to 2.13, reducing the KLR false-positive rate to 8%. However, normal KLR persisted among 16% (14/85) of samples with free λ chains by immunofixation, warranting caution during interpretation.

Conclusions

Our analysis indicated that revision of Freelite SFLC reference intervals improves assay interpretation and should prompt reconsideration of Freelite reference intervals worldwide.

Key Points.

• International guidelines for multiple myeloma management use the Binding Site Freelite serum free light chain (SFLC) assay’s manufacturer-recommended reference intervals to define “abnormal.”

• Among 5057 patients with monoclonal-negative results by both SFLC and immunofixation assays, 25% had κ/λ ratios outside the manufacturer-defined normal reference interval of 0.26-1.65.

• Manufacturer-recommended normal reference intervals for the Freelite SFLC assay warrant reevaluation and revision; a κ/λ ratio reference interval of 0.67-2.13 was most appropriate for our population tested on the Optilite platform.

INTRODUCTION

Monoclonal gammopathies comprise a spectrum of disorders that range from asymptomatic monoclonal gammopathy of undetermined significance (MGUS) to non-Hodgkin lymphoma, lymphoplasmacytic lymphoma, multiple myeloma, and amyloid light chain (AL) amyloidosis. The more advanced of these conditions are typically characterized by overproduction of clonal immunoglobulins that may be identified as intact immunoglobulins (immunoglobulin G [IgG], IgA, IgM, IgD, or IgE linked to κ or λ light chains), or free κ or λ light chain molecules without accompanying heavy chains. Serum and urine protein electrophoresis (PE), immunofixation, and serum free light chain (SFLC) assays are used clinically to identify and quantify disease burden, and they provide critical information for patient screening, diagnosis, and treatment response assessment.

For many years following initial US Food and Drug Administration (FDA) approval in 2001, the Freelite assay (Thermo Fisher Scientific) was the sole method for clinical measurement of SFLC. This quantitative assay employs polyclonal antisera to detect κ and λ light chain epitopes that are exposed when light chains are not bound to heavy chains. The manufacturer-recommended normal reference intervals used data from a study published in 2002 and included 100% of the FLC results from 282 healthy donors: 0.33 to 1.94 mg/dL for κ FLC (KFLC), 0.57 to 2.63 mg/dL for λ FLC (LFLC), and 0.26 to 1.65 for the κ/λ ratio (KLR).¹ Inclusion of all values from this healthy cohort yielded “diagnostic ranges” that were wider than typical reference intervals, which typically include only the central 95% of values from a healthy population. As the manufacturer continues to promote these diagnostic ranges as the recommended normal reference interval, we refer to them in this report as “manufacturer-recommended reference intervals.” It is standard practice for these intervals to be verified in individual clinical laboratories in accordance with College of American Pathology guidelines.

Multiple studies established the Freelite SFLC assay’s increased sensitivity over serum and urine immunofixation for the detection of FLC and validated the clinical utility of quantifying these molecules.^1-5 Consequently, SFLC testing using this assay is in widespread clinical use. Many clinical guidelines for the care of patients with plasma cell neoplasms (PCNs), including those from the International Myeloma Working Group (IMWG), have provided specific recommendations for the use and interpretation of results from the Freelite SFLC assay.^5,6 These recommendations include using SFLC results to screen for a variety of PCNs, including AL amyloidosis,^5,6 and to determine prognosis for MGUS,⁷ solitary plasmacytoma,⁸ and smoldering multiple myeloma,⁹ where values outside the manufacturer-defined KLR thresholds are associated with a greater risk of progression to myeloma. Additional recommendations use normal KLRs to define a complete response to treatment among patients with AL amyloidosis and a stringent complete response to therapy in patients with multiple myeloma.^10,11 Together, these clinical guidance documents underscore the importance of Freelite SFLC results for the clinical care of patients with PCNs and of the manufacturer-defined KLR reference interval in particular.

The goal of this study was to retrospectively evaluate the clinical performance of the Freelite SFLC assay and to assess the accuracy of SFLC compared with immunofixation results. As a Clinical Laboratory Improvement Amendments (CLIA)–certified regional reference laboratory associated with a major cancer center, we performed more than 95,000 PE, immunofixation, and SFLC assays from 2010 through 2020 using FDA-approved assays and instrumentation. These clinical data were comprehensively analyzed in the current study, which includes a reevaluation of the Freelite assay’s normal reference intervals for KFLC, LFLC, and KLR. Our investigation has yielded compelling results that we expect will affect patient diagnosis and care.

METHODS

Data Generation

Serum and urine PE and immunofixation were performed on consecutive inpatient and outpatient clinical serum samples at our institution from January 2010 to December 2020 using Sebia HYDRASYS or HYDRASYS 2 Scan instruments with Phoresis software (Sebia), according to the manufacturer’s instructions. Immunofixation interpretations were performed by 8 experienced board-certified MD- or PhD-level experts who regularly participated in competency evaluations and informal consultations to ensure consistency in interpretations. Clinical information was typically not available at the time immunofixation interpretations were performed. Over the study period, immunofixation interpretations were considered the gold/reference standard for the identification of monoclonal components in serum, to which SFLC results were compared. The SFLC results were generated using the Freelite turbidimetric assay on the BN II (May 2006-March 2014; Siemens Healthineers), SPA (March 2014-June 2017; Binding Site/Thermo Fisher Scientific), and Optilite (June 2017-current; Binding Site/Thermo Fisher Scientific) platforms according to manufacturer protocols. Assay correlations between BN II and SPA instruments and between SPA and Optilite instruments were performed using clinical specimens, with correlation coefficients for KFLC, LFLC, and the KLR ranging from 0.92 to 0.99. Specifically, the KLR correlation between the BN II and SPA instruments was 0.968, with a Deming regression slope of 1.02 and y-intercept of 0.11, and with a Passing Bablok regression slope of 1.01 and y-intercept of 0.04. Between the SPA and Optilite instruments, the KLR correlation was 0.995, with a Deming regression slope of 0.85 and y-intercept of −0.10 and with a Passing Bablok regression slope of 0.82 and y-intercept of −0.03. The average bias between the BN II and SPA instruments and between the SPA and Optilite instruments was found to be 0.14 (or 19%) and −0.47 (or −15%), respectively. The SFLC results were reported independently of electrophoresis and immunofixation results. Data were obtained from our data warehouses, which contain laboratory orders and results as well as inpatient and outpatient encounters. All activities, including patient medical chart review, were covered by an institutional review board–approved protocol.

Study Population

Our Clinical Immunology Laboratory serves an urban medical center hospital network that includes an academic medical center, a central-city public safety-net hospital, a designated cancer center, and a regional reference laboratory population of rural and urban individuals; our 5-state catchment area consists of 11.6 million people (2022 US Census Bureau data), but most samples come from the nearly 7.8 million residents of our state. According to 2022 census data, 66% of the state’s residents identified as non-Hispanic White, while 13.7% identified as Hispanic or Latino, 10% as Asian, 5.2% as two or more races, 4.5% as Black or African American, 2% as American Indian or Alaska Native, and 0.8% as Native Hawaiian or other Pacific Islander. In addition, 16.2% of state inhabitants were older than 65 years of age, compared with the national average of 16.8%. The 2015-2019 average annual incidence rate of multiple myeloma was 6.6 per 100,00 individuals, slightly lower than the national average of 7 per 100,000 (American Cancer Society). The United Health Foundation’s America’s Health Rankings 2022 Annual Report ranked our state as the ninth healthiest overall, and sixth, 11th, and 13th for health behaviors, clinical care, and health outcomes, respectively.

Analysis Software and Approaches

Data were analyzed in RStudio (Posit) using R, version 3.6.1 (R Foundation for Statistical Computing).^12,13 The tidyverse, janitor, and lubridate packages were used to tidy data before analysis.^14-16 The methods for data tidying, including inclusion and exclusion of specimens and annotation with clinical and International Statistical Classification of Diseases (ICD) data, are described in the sections that follow and in our Supplementary Materials (all supplementary material is available at American Journal of Clinical Pathology online). The HDInterval package was used to calculate highest-density intervals containing 50% of data (credMass = 0.5), while the stats package was used to calculate quantiles/percentiles.¹⁷ Estimates and bootstrapped 95% CIs for median yearly KLR values were estimated using the R package confintr (boot_type = “bca,” R = 9999).¹⁸ Indirect reference intervals were estimated from study data using the RefineR package,¹⁹ while indirect reference intervals were established using the package referenceIntervals, using a nonparametric approach to define the central 95% of the data.^19,20 Detailed methods for defining these indirect reference intervals are described below. The packages tidyverse, cowplot, and gridExtra were used to generate graphics, while ggplot’s geom_quantile function (quantiles = 0.5, method = “rq”) was employed to plot median quantile regression lines.^14,21,22

Specimen Categorization and Inclusion/Exclusion Criteria

Sample Selection by Immunofixation Result

The time period for sample inclusion was chosen randomly, but with the goal of maximizing the number of samples that met our eligibility criteria; no sample size calculation was performed. Therefore, all consecutive clinical samples with a PE or immunofixation result reported between 2010 and 2020 were eligible for inclusion. Samples were categorized by the presence or absence of a single monoclonal component, according to the reported immunofixation interpretation result, outlined in Figure 1 and in the Supplementary Materials. Only “first-occurrence” specimens (ie, the first time a particular monoclonal component was identified for that patient) were included in the analyses to eliminate the effects of repeated testing of individuals with known monoclonal components and to limit bias introduced when a patient’s previously identified monoclonal component was already known. Samples with indeterminate oligoclonal banding were not included in our data set. All quality control, research, and proficiency testing specimens were excluded from the analyses, as were specimens without medical record number information Figure 1.

FIGURE 1.

Data analysis flowchart and descriptive characteristics of included specimens. Serum electrophoresis/immunofixation and SFLC data from 95,609 specimens tested over a 10-year period were filtered to yield a subset of 8178 first-occurrence specimens with concurrent SFLC results for analysis. Specimens were grouped according to their respective immunofixation interpretations, and descriptive characteristics of patient sex and age from which these samples derive are shown. Patients for whom sex was unknown (n = 7) were included in the count for total number of specimens. This figure was created with BioRender.com. eGFR, estimated glomerular filtration rate; ICD International Statistical Classification of Diseases; Ig, immunoglobulin; κMFLC, κ monoclonal free light chain; λMFLC, λ monoclonal free light chain; SFLC, serum free light chain.

Integration of SFLC Data

For each patient, every first-occurrence specimen was paired with an SFLC result obtained within 2 weeks of the serum PE or immunofixation collection date. First-occurrence specimens without a paired SFLC result were excluded from subsequent analysis.

Assessment of Patient Kidney Function and Urine SFLC Results

The estimated glomerular filtration rate (eGFR) was calculated as GFR in mL/min/1.73 m² by the Modification of Diet in Renal Disease study equation until June 1, 2020, when eGFR was calculated using the 2009 Chronic Kidney Disease (CKD) Epidemiology Collaboration_{NO RACE} equation.²³ All specimens with eGFR data obtained within 2 weeks of the PE or immunofixation serum sample collection date were identified. First-occurrence specimens were annotated with their respective National Kidney Foundation eGFR category when available. All patients in whom Bence-Jones protein had ever been detected were also identified and annotated as being Bence-Jones protein positive (see Supplementary Materials).

Patient Diagnosis Assignments

Monoclonal component–related diagnoses were identified using available encounter-level ICD-9 or ICD-10 codes from outpatient lab orders as well as outpatient and inpatient encounters (see Supplementary Materials). The ICD-coded events were captured if they were issued up to 5 days before or 2 days after the serum PE or immunofixation collection time. This time frame was selected to capture ICD codes relevant to the serum PE or immunofixation order. We did not have access to primary medical diagnoses. To capture individuals with monoclonal FLCs by immunofixation who were subsequently diagnosed with a monoclonal gammopathy and thus assigned relevant ICD codes outside this 7-day window, we performed comprehensive chart review for all 323 patients with monoclonal FLCs to ensure correct assignment of disease categories.

Binomial Multivariate Regression

Binomial logistic regression was performed using the generalized linear modeling (glm) function from the stats R package to predict KLR discordance using age, sex, eGFR, and predefined patient diagnosis categories. The model was iteratively refit, and χ² tests of the difference in deviance between models with and without each covariate were performed. To test for the effect of eGFR, specimens without eGFR data were excluded. Only diagnoses of PCN and amyloidosis yielded statistically significant deviance in these analyses. The regression model was therefore refit using these 2 covariates, which were found to be significant predictors of concordance.

Orthogonal SFLC Quantitation

The Sebia SFLC assay, an enzyme-linked immunoassay (ELISA) that employs polyclonal rabbit antisera to detect free KFLC and LFLC,^24,25 was used to test selected samples in an external CLIA-certified laboratory using deidentified specimens.

Monoclonal-Negative Specimen Selection for Reference Interval Analyses

First-occurrence monoclonal-negative specimens for initial reference interval analyses included a subgroup of 1537 specimens, tested using 3 different platforms over the study period. We did not include monoclonal-negative patients with different stages of kidney disease in these reference cohorts. Charlson comorbidities²⁶ as well as Nicholson and Fortin conditions²⁷ were both assigned to this monoclonal-negative population (n = 1537) using the R package multimorbidity.²⁸ The Charlson comorbidities consist of 17 inpatient medical conditions that can be differentially weighted and summed to generate a score that predicts mortality in longitudinal studies. The Nicholson and Fortin conditions comprise 20 chronic categories of conditions that have high relevance, prevalence, and impact to primary care. We did not distinguish between inpatient and outpatient ICD codes and required that at least 2 codes match a category for a comorbidity or condition to be positively coded. Subsequently, a final monoclonal-negative reference group was defined by limiting SFLC results to those generated using the Optilite platform (n = 1199), as described in Figure 1 and in the Supplemental Methods.

Current Reference Interval Thresholds

In accordance with the College of American Pathology guidelines for evaluation of FDA-approved in vitro diagnostics, our laboratory verified manufacturer-recommended reference ranges in 2006, 2014, and in 2016, when the Freelite assay was transitioned from the BN II instrument to the SPA and Optilite platforms, respectively. Unless otherwise stated, the manufacturer-recommended KLR reference interval of 0.26 to 1.65 was used for categorizing samples as having abnormally high (>1.65), normal, or abnormally low (<0.26) KLRs. Samples with no monoclonal component are expected to possess a normal KLR. Those with κ monoclonal FLC (KMFLC) identified by immunofixation are expected to demonstrate abnormally high KLR, while those with λ monoclonal FLC (LMFLC) by immunofixation should demonstrate an abnormally low KLR.

New Reference Interval Thresholds

Indirect reference intervals in this study were established from a nondiseased Optilite-tested “reference” population, based on serum and urine immunofixation, eGFR results, and ICD codes Figure 1 (Supplemental Methods). Reference intervals were calculated with the package referenceIntervals using a nonparametric approach to define the central 95% of the data (reference interval = “n”, refConf = 0.95).²⁰ We calculated 90% CIs using the package’s basic bootstrap method (CI = “boot,” limitConf = 0.90, bootStat = “basic”). When specified, this newly defined reference interval of 0.67 to 2.13 was used for categorizing samples as having abnormally high (>2.13), normal, or abnormally low (<0.67) KLRs.

KLR Reference Interval Estimation

The refineR algorithm,¹⁹ which indirectly estimates reference intervals from real-world data, was used to independently estimate appropriate reference intervals. This algorithm performs iterative, regularized maximum likelihood optimization analyses on Box-Cox–transformed normal distributions of the data to best model the underlying diseased and nondiseased distributions that are both present in the data. The algorithm estimates a model of the nondiseased distribution, and a reference interval is derived from its 2.5th and 97.5th percentiles. A 95% CI of these estimated reference interval limits was computed from 200 bootstrap iterations, and the median values of these bootstrap estimates were used as the estimates for the lower and upper reference interval thresholds (pointEst = c(“medianBS”), NBootstrap = 200).

RESULTS

To explore SFLC assay performance, we obtained SFLC and PE/immunofixation data from 95,609 specimens collected from 43,971 individuals submitted to our laboratory during a 10-year period Figure 1. All first-occurrence specimens with a single monoclonal component and concurrent SFLC results were identified, yielding 8178 samples for analysis. The percentage of specimens from male and female individuals and their ages were stable across immunofixation interpretation groups and consistent with the known epidemiology of MGUS, multiple myeloma, and AL amyloidosis.^29-31

We simplified our data set by grouping specimens with intact heavy and light chains into 2 categories according to light chain expression: IgK and IgL. Of the 3119 total monoclonal component–containing specimens, 1827 (59%) bore κ light chains, while 1292 (41%) bore λ light chains Figure 1 (bottom box). In addition, 323 (10.4%) of the monoclonal-positive specimens were classified as having κ or λ monoclonal FLC only by immunofixation (KMFLC or LMFLC, respectively). These data are consistent with the known higher frequency of κ vs λ light chain involvement in monoclonal gammopathies and the overall frequency of light chain disease.³⁰ Consistent with a previous report,³² a substantial percentage of IgK and IgL samples had normal FLC values Figure 2. As expected for abnormal immunofixation-defined KMFLC specimens, the vast majority of the KFLC values fell well above the reference interval threshold of 1.94 mg/dL Figure 2A. Similarly, for abnormal immunofixation-defined LMFLC specimens, the majority of LFLC values fell above the reference interval threshold of 2.63 mg/dL Figure 2B.

FIGURE 2.

Distribution of serum free light chain (FLC) values from 8178 unique samples according to contemporaneous immunofixation results. A, B, Violin plots of the population distribution (marginal density estimates) of the log-transformed κ FLC values (A) and λ FLC values (B) are plotted by immunofixation interpretation (x-axis). The manufacturer-defined reference intervals for κ FLC (0.33-1.94 mg/dL) and λ FLC (0.57-2.63 mg/dL) are represented as black dashed horizontal lines. Each violin plot is overlaid with a quantile box plot of the data in which the lower whisker extends to the 1st percentile while the lower boundary of the box is formed by the 25th percentile. The crossbar indicates the median, while the upper boundary of the box is formed by the 75th percentile, with the upper whisker extending to the 99th percentile. The horizontal dashes across the lower and upper whiskers correspond to the 2.5th and 97.5th percentiles, respectively, and are annotated by the corresponding antilog value. Ig, immunoglobulin; κMFLC, κ monoclonal free light chain; λMFLC, λ monoclonal free light chain.

Concordance Between Immunofixation and Serum FLC Results Is Superior for KMFLC-Containing Samples Compared With LMFLC-Containing Samples

Since the KLR calculation is key for establishing the clonality of a FLC elevation, we have primarily focused the remainder of this report on KLR values. Samples containing immunofixation-defined KMFLC should exhibit abnormally high KLR, whereas those containing LMFLC should exhibit abnormally low KLR (both defined as “concordant” results). When we assessed concordance between immunofixation interpretation and KLR categorization for KMFLC-containing samples using current manufacturer-recommended thresholds, the concordance was 98%, with 2% exhibiting normal KLR results Table 1 (Supplemental Figure S2).

TABLE 1.

Concordance of immunofixation interpretation with serum free light chain KLR categories using the manufacturer-defined KLR reference interval^a

Immunofixation interpretation	Normal KLR, %	Concordant KLR, %	Discordant KLR,b %
KMFLC (n = 127)	2	98	0
LMFLC (n = 196)	15	82	3
IgK (n = 1701)	34	66	0
IgL (n = 1097)	66	25	9
Monoclonal negative (n = 5057)	75	—	25

Ig, immunoglobulin; KLR, κ/λ ratio; KMFLC, κ monoclonal free light chain; LMFLC, λ monoclonal free light chain.

^aKLR from first-occurrence specimens (n = 8178) were categorized by immunofixation result (left column) and evaluated for abnormally high, normal, or abnormally low KLR results using the current manufacturer-defined reference interval (0.26-1.65).

^b“Discordant” refers to samples with high KLR values in samples with IgL or LMFLC by immunofixation, low KLR values in samples with IgK or KMFLC by immunofixation, or abnormal KLR values in samples that were monoclonal negative.

By contrast, the 196 LMFLC-containing specimens demonstrated concordance of only 82% Table 1. The 35 LMFLC specimens with normal or discordant KLR did not segregate by the physician immunofixation interpreter of record or by the month or year of collection (data not shown). Of the 30 LMFLC-containing specimens with normal KLR, 19 (63%) exhibited absolute LFLC and KFLC values above their respective upper reference interval thresholds (Supplemental Figure S2), resulting in a normal KLR upon calculation. Among the 5 LMFLC-containing specimens with high KLR, the KLRs ranged from 1.66 to 1.74 (Supplemental Figure S2).

Poorer Concordance Between Immunofixation and Serum FLC Results for LMFLC Compared With KMFLC Is Not Explained by Kidney Dysfunction or Disease Diagnosis

In patients with CKD, SFLC levels can increase as both filtering and resorptive capacities decline. This, in turn, can increase the KLR in the absence of a monoclonal gammopathy.^32-36 In patients with LMFLC identified by immunofixation, impaired kidney function could cause KFLC elevations that offset the monoclonal-driven rise in LFLC, yielding LMFLC samples with normal KLR. We therefore investigated whether exclusion of specimens from individuals with decreased kidney function (measured by eGFR) restored the ability of the SFLC assay to accurately detect LMFLC identified by immunofixation Figure 3. In total, 15.3% (30/196) of LMFLC-containing specimens demonstrated normal KLR Figures 3A-3B. When all specimens with concurrent eGFR less than 60mL/min/1.73m² were excluded, 15.4% (22/143) of LMFLC-containing specimens continued to demonstrate a normal KLR Figures 3C-3D. This lack of change suggested that factors other than kidney dysfunction underlie this discrepant assay performance.

FIGURE 3.

Serum free light chain KLRs according to immunofixation result before and after exclusion of patients with poor kidney function. A, Log-transformed KLR values from individual samples are plotted by immunofixation interpretation (x-axes) and colored by KLR interpretation according to the manufacturer-recommended KLR reference interval (0.26-1.65; black dashed horizontal lines) (A). The number of samples in each immunofixation group are as listed in Figure 1 (monoclonal-negative, n = 5057; IgK, n = 1701; KMFLC, n = 127; IgL, n = 1097; LMFLC, n =196). B, For each immunofixation interpretation group, the proportion of specimens with abnormally high, normal, and abnormally low KLR is shown. C, D, All patients with compromised kidney function, defined as an estimated glomerular filtration rate below 60 mL/min/1.73m², had their samples excluded from the data set, and the KLR data were reanalyzed similar to (A) and (B). The number of samples in each immunofixation group after exclusion of patients with decreased kidney function are monoclonal-negative, n = 4021; IgK, n = 1343; KMFLC, n = 80; IgL, n = 876; and LMFLC, n = 143. Statistical significance between the KMFLC and LMFLC populations was calculated comparing the proportions of KLR-concordant and KLR-discordant/normal specimens using Pearson χ² test with Yates continuity correction. *P < .01; **P < .0001. Ig, immunoglobulin; KLR, κ/λ ratio; κMFLC, κ monoclonal free light chain; λMFLC, λ monoclonal free light chain.

We further sought to understand whether differences in underlying disease diagnoses could contribute to the LFLC discordance we observed. For the 323 individuals with serum monoclonal FLCs detected by immunofixation, disease conditions were evaluated using available encounter-level ICD-9 or ICD-10 codes (see “Methods”) and comprehensive chart review Figure 4. The majority of KMFLC- and LMFLC-bearing patients (defined by immunofixation) were found to carry a biopsy-proven diagnosis of PCN, with or without CKD or AL amyloidosis (pink shades). A smaller number of patients had CKD, amyloidosis, leukemia, or lymphoma Figure 4A-4B. LMFLC-positive specimens with normal KLR were found within every disease category Figure 4B, 4D, 4F. In addition, using multivariate binomial regression, we found that covariates of age, sex, concurrent eGFR, or disease category were not significant predictors of discordance between the SFLC and immunofixation results for the detection of LFLCs.

FIGURE 4.

Immunofixation and serum free light chain results contextualized by disease diagnosis. A, B, Patients with first-occurrence KMFLC (A) and LMFLC (B) by immunofixation were classified by clinical diagnosis (x-axis) and the number of individuals in each diagnosis group plotted (y-axis, A, B). Stacked histograms display the total number of individuals with a given diagnosis (PCNs [pink shades]; CKD [orange shades]; amyloidosis [blue shades]; leukemia [dark purple]; lymphoma [light purple]; or other/no ICD codes [gray shades]). The total number of individuals within the diagnosis group is shown above each histogram column; each individual is counted once within a column but may be counted again within another grouped column if that patient carries multiple diagnoses. C, D, Log-transformed KLR values for KMFLC (C) and LMFLC (D) specimens are plotted by diagnosis group and colored by diagnosis, consistent with colors used in A and B. The manufacturer-recommended KLR reference interval (0.26-1.65) is shown as black dashed horizontal lines. E, F, For each diagnosis group, the proportion of KMFLC (E) and LMFLC (F) specimens that have abnormally high, normal, and abnormally low KLRs according to manufacturer-defined KLR reference thresholds is shown. CKD, chronic kidney disease; ICD, International Statistical Classification of Diseases; KLR, κ/λ ratio; κ monoclonal free light chain; LMFLC, λ monoclonal free light chain; PCN, plasma cell neoplasm.

Chart review of the 30 λ-discordant patients with LMFLC by immunofixation and normal KLR (0.26-1.65) results revealed that 5 had no evaluable records and 8 had no information about light chain disease from our institution. Two patients had documented κ light chain disease by flow cytometry, and 7 had CKD without evidence of a PCN, raising the possibility that the small λ light chains detected by immunofixation represented oligoclonal bands. Among 8 of 25 evaluable patients, however, λ light chain disease was documented by an orthogonal method, including urine immunofixation and λ restriction by flow cytometry or immunohistochemistry. This finding suggests that in at least a subset of cases, the SFLC assay failed to identify underlying disease that was captured by immunofixation.

To investigate potential explanations for the detection of small LMFLC by the immunofixation but not SFLC assay, we tested 6 recently identified λ-discordant specimens. We performed 10-fold and 100-fold dilutions of the discordant sera to rule out a high-dose hook effect in the Freelite assay. To investigate whether an alternative polyclonal antiserum could detect these patients’ LMFLC when the Freelite antisera could not, discordant sera were tested using an orthogonal ELISA-based SFLC assay (Sebia). The Sebia assay, however, did not identify abnormally low KLR in any of the 6 Freelite-based discordant specimens. We also performed immunofixation using anti-IgD and anti-IgE antisera to exclude the possibility that the LMFLCs visualized on immunofixation were instead IgDL or IgEL monoclonal components. Finally, immunofixation was repeated using diluted serum to improve signal-to-noise ratios, which confirmed the presence of small monoclonal free λ bands (data not shown).

A Substantial Proportion of Monoclonal-Negative Specimens Yield Serum FLC Levels Above Manufacturer-Defined Normal Reference Intervals

Through our comprehensive analysis of SFLC assay performance, we unexpectedly found that the overall distribution of KLR values from our first-occurrence data set was skewed to the right of the manufacturer-defined reference interval (indicated by the vertical dashed lines) Figure 5A-5B). Moreover, an abnormal KLR was present in one-quarter of monoclonal-negative specimens Figure 3A, 3B and Table 1. Using the refineR algorithm, which applies statistical methods to model the distribution of results from laboratory data and estimate reference intervals,¹⁹ the normal KLR reference interval was found to be 0.68 to 2.44, substantially higher than the manufacturer-recommended reference interval of 0.26 to 1.65. We further investigated the distribution of KLR values reported at our institution over the duration of this study (2010-2020). For all specimens with reported SFLC results (n = 48,012), more than 70% of KLR values were consistently higher than the original reference interval median KLR of 0.6 (Supplemental Figure S3A).¹ Among first-occurrence specimens, this shift was more pronounced (n = 8178 and n = 5057) Figure 5C (Supplemental Figure S3B-3C). Together, these findings suggested that the current KLR reference interval does not accurately describe our population.

FIGURE 5.

Distribution of first-occurrence sample KLR values, analyzed over time and by assay platform. A, B, The smoothed histogram of the kernel density estimate for log-transformed KLR values is plotted for all 8178 first-occurrence specimens, analyzed in aggregate (A) and by immunofixation interpretation (B). The manufacturer-defined KLR reference interval (0.26-1.65) is shown as black dashed vertical lines. C, Percentiles of log-transformed KLR values generated from January 2010 through December 2020 are plotted monthly over time for all first-occurrence specimens (n = 8178). The 2.5th, 5th, 25th, 50th, 75th, 95th, and 97.5th percentiles are shown. A marginal histogram representing the number of specimens that contribute to the distribution at any given month is plotted along the top of each graph. Months with more than 20 results are shaded gray. The manufacturer’s KLR reference interval (0.26-1.65) is represented by black dashed horizontal lines, while the originally reported median KLR value of 0.6 (1) is represented by a gray dashed horizontal line. D, Violin plots of the distribution (marginal density estimates) of KLR values are plotted by year and platform for all first-occurrence samples in our study (n = 8178). Each violin plot is overlaid with a quantile box plot of the data, as in Figure 2. The estimated median and its 95% bootstrapped CI are shown above each violin plot, while the number of tested samples is shown below. Ig, immunoglobulin; KLR, κ/λ ratio; κMFLC, κ monoclonal free light chain; λMFLC, λ monoclonal free light chain; NOMCO, monoclonal-negative.

Since our results were generated over a decade using 3 different platforms, we examined KLR distributions of first-occurrence specimens, grouped by year and contextualized by the platform used to generate the results Figure 5D. These analyses demonstrated that median yearly KLR values generated on the BN II instrument (2010-2014) fell well above the original reference interval median of 0.6 (gray dashed line) and also increased over time. Although the median KLR values were similar before and shortly after transitioning from the BN II to the SPA instrument, an upward trend in medians became quite prominent between 2015 and 2017, a period when testing volumes also increased substantially. Overall, the SPA instrument was found to have a 19% bias relative to the BN II, while the Optilite platform demonstrated a −15% bias relative to the SPA instrument (see “Methods”). The transition to Optilite in 2017 was accompanied by a downward adjustment of the median, although KLR values gradually increased again by 2020. These trends could also be visualized when KLR values from the first-occurrence and monoclonal-negative specimens were plotted over time without log transformation (Supplemental Figure S4B and S4C, respectively). Between 2010 and 2020, a small and gradual increase in median KLRs was observed in individuals younger than 50 years of age. These data suggested a contribution of assay drift because this population is less likely to have increasing KLR related to aging or decreased kidney function (Supplemental Figure S4B). Finally, the distribution of patient ages remained relatively stable over the course of the study, with the median age increasing slightly by 3 years, from 62 in 2010 to 65 in 2020 (data not shown). Together, these observations point to multiple factors, including analytical bias and assay drift, that may have contributed to changes in the distribution of KLR values over time. Most importantly, however, the median KLR from our population was consistently above the original reference interval median of 0.6 from the start of the study period, indicating that the manufacturer-defined SFLC reference intervals have not been appropriate for our population for many years.

Revised SFLC Reference Intervals Improve the Classification of Specimens From Both Monoclonal–Negative and Monoclonal-Positive Populations

To define new reference intervals, we first identified a population of 1537 monoclonal-negative individuals without disease known to increase SFLC Figure 1(Supplemental File 1). Individuals who ever had any monoclonal components in the serum or urine were excluded as were those with no available ICD codes or a contemporaneous eGFR less than 60mL/min/1.73m². Most individuals (84% [n = 1289/1537]) in this monoclonal-negative population did not possess ICD codes related to any Charlson (inpatient-relevant) comorbidities, while 56% (861/1537) of individuals had no Nicholson and Fortin (primary care–relevant) conditions assigned. Of those with Nicholson and Fortin conditions, the most common diagnoses were musculoskeletal (276/1537 [18%]), coronary vascular disease (133/1537 [8.7%]), and hypertension (131/1537 [8.5%]).

We then further restricted the monoclonal-negative population to results generated on the Optilite platform (n = 1199). The median age of this Optilite-only monoclonal-negative reference (61 years for men, 59 years for women) was slightly lower than that of the overall study population (63 years for men, 63 years for women) Figure 1. Within this monoclonal-negative reference population, 12% possessed abnormally high KLR using the manufacturer-recommended threshold (Supplemental Figure S5A, top panel), consistent with the false-positive rate described among healthy volunteers.³⁷ The SFLC results from the Optilite platform were characterized by stable KLR distribution percentiles across time and reagent lots Figure 6A. With age, the KLR distribution median increased modestly in this population Figure 6B, likely related to normal ageing-associated decreases in kidney function.³⁶ Regardless of assay date or patient age, even among patients younger than 50 years of age, all observed KLR distributions consistently included a substantial proportion of specimens with values above the upper limit of the manufacturer-recommended reference interval.

FIGURE 6.

Serum free light chain values from 1199 Optilite-tested patients with no monoclonal component detected by immunofixation, analyzed by sample collection date and patient age. A, B, KLR percentiles are plotted from the Optilite-only reference monoclonal-negative population (n = 1199) and trended by collection date (A) and patient age (B). The 2.5th, 5th, 25th, 50th, 75th, 95th, and 97.5th percentiles are shown. A marginal histogram representing the number of specimens that contribute to the distribution at any given date or age is plotted along the top of each graph. Months with more than 20 results are shaded gray. The manufacturer’s KLR reference interval (0.26-1.65) is shown as black dashed horizontal lines, with the originally reported median KLR value of 0.6 (1) shown as a gray horizontal dashed line. Dates at which new κ or λ serum free light chain assay reagent lots were introduced are shown as gray dashed vertical lines. C-E, Scatter and violin plots of KFLC (C), LFLC (D), and KLR (E) values derived from the Optilite-only reference monoclonal-negative population are shown. The highest-density interval that contains 50% of the data is indicated by vertical black bars to the left of each scatter plot. Each violin plot is overlaid with a quantile box plot of the data, as in Figure 2. The horizontal dashes across the lower and upper whiskers correspond to the 2.5th and 97.5th percentiles, respectively, and are annotated by the corresponding antilog value representing the proposed new reference interval thresholds. The 90% CI (n = 5000 bootstraps) is shown in parentheses next to the respective reference interval threshold. KLR, κ/λ ratio; κMFLC, κ monoclonal free light chain; λMFLC, λ monoclonal free light chain.

In keeping with the original manufacturer-recommended reference interval, we identified a diagnostic range using 100% of results from our Optilite-only monoclonal-negative reference population (n = 1199) (Supplementary Table 2). We then recategorized all first-occurrence Optilite samples (n = 4601) using this diagnostic range (Supplementary Table 3) and evaluated the concordance of Optilite SFLC results with immunofixation interpretations. Discordant KLRs were essentially eliminated for LMFLC-containing samples, but those containing KMFLC demonstrated an increase in the number of falsely negative KLRs from 2% to 8%. Also, for samples with IgK monoclonal components, a substantial increase in the percentage of KLR-normal results was observed. We therefore calculated the central 95% data interval (2.5th-97.5th percentiles) and 90% CIs for KFLC, LFLC, and the KLR from the Optilite-only monoclonal-negative reference population Figure 6C-6E, yielding new reference intervals of 0.81 to 6.05 mg/dL for KFLC, 0.70 to 4.66 mg/dL for LFLC, and 0.67 to 2.13 for the KLR.

To evaluate the consequences of using our newly defined KLR reference interval, all first-occurrence sample results generated on the Optilite platform (n = 4601) were categorized using the original and Optilite-only KLR reference intervals Table 2 (Supplemental Figure S6). The KLR classification (abnormal high, normal, abnormal low) of samples differed between these 2 reference intervals, and the new reference interval decreased the number of monoclonal-negative specimens with an abnormal KLR from 718 (22.4%) to 267 (8.3%) Table 2 (Supplemental Figure S5C and Supplemental Figure S7). The KLR classification of KMFLC specimens remained unchanged, while 3 LMFLC-bearing individuals with abnormally high KLRs were recategorized as having a normal KLR, and 2 LMFLC-bearing individuals with normal KLRs were recategorized as having an abnormally low KLR. Among the 2 λ cases that were reclassified from KLR-normal to KLR-low, chart review revealed that 1 individual had biopsy-proven λ restriction of their PCN, while the other had marginal zone lymphoma but no documented light chain restriction by flow cytometry or immunohistochemistry. It is worth noting that the application of the new reference interval did not eliminate the inability of the SFLC assay to identify a small subset of LMFLC-containing samples detected by immunofixation Table 2. When KLR results fall close to reference interval thresholds, absolute FLC values can be used in conjunction with the KLR to detect disease. Changes to our KFLC and LFLC thresholds either did not affect or improved the accuracy of KLR classification among samples whose FLC categories changed (Supplementary Table 4). Thus, the new FLC thresholds modestly improved our ability to identify abnormal KMFLC- and LMFLC-containing samples based on the KLR. Overall, the new reference interval reduced the false-positive rate in monoclonal-negative samples and improved the classification of LMFLC specimens.

TABLE 2.

Concordance of immunofixation interpretation with serum free light chain KLR categories using manufacturer-defined and revised KLR reference intervals for samples with Optilite-generated serum free light chain results^a

Immunofixation interpretation	Normal KLR, %	Concordant KLR, %	Discordant KLR,b %
Manufacturer-defined KLR reference interval (0.26-1.65) c
KMFLC (n = 51)	2	98	0
LMFLC (n = 85)	15	81	4
IgK (n = 729)	35	65	0
IgL (n = 519)	70	21	9
Monoclonal negative (n = 3217)	78	—	22
New KLR reference interval (0.67-2.13) c
KMFLC (n = 51)	2	98	0
LMFLC (n = 85)	16	84	0
IgK (n = 729)	49	50	1
IgL (n = 519)	58	39	3
Monoclonal negative (n = 3217)	92	—	8

Ig, immunoglobulin; KLR, κ/λ ratio; KMFLC, κ monoclonal free light chain; LMFLC, λ monoclonal free light chain.

^aOnly first-occurrence samples with serum free light chain results generated on the Optilite platform (n = 4601) are included in this table.

^b“Discordant” refers to samples with high KLR values in samples with IgL or LMFLC by immunofixation, low KLR values in samples with IgK or KMFLC by immunofixation, or abnormal KLR values in samples that were monoclonal negative.

^cKLR from these Optilite-tested specimens (n = 4601) were categorized by immunofixation result (left column) and evaluated for abnormally high, normal, and abnormally low KLR using the current manufacturer-defined reference interval (top) and the new proposed reference interval (bottom).

To confirm these new SFLC assay thresholds, we applied them to an alternative population consisting of 914 patients with normal serum electrophoresis patterns whose samples were also tested on the Optilite platform. These specimens were not tested by immunofixation either because the clinical provider did not request it or because immunofixation was not indicated based on the electrophoresis result. Using the manufacturer-recommended thresholds, 24% of this population had an abnormally high or low KLR. Upon implementation of the new KLR reference interval, only 9.6% had abnormal KLRs (Supplemental Figure S5D), demonstrating that the new reference thresholds reduced the percentage of samples with abnormal KLRs in the normal electrophoresis pattern specimens, as it did for the monoclonal-negative specimens.

Finally, to formally verify our new reference intervals for KFLC, LFLC, and KLR, we tested 106 specimens recently collected from 52 male and 54 female healthy blood donors. The KFLC and LFLC values ranged from 0.75 to 3.43 and 0.63 to 3.28 mg/dL, respectively. The resultant KLR values ranged from 0.78 to 2.91, and the central 95% of results (2.5th-97.5th percentiles) spanned from 0.85 to 1.92. Among these normal specimens from healthy adults, 12.3% (13/106) displayed KLR values above the current upper threshold of 1.65, while only 12 specimens had a KLR below 1.00. The median age of these donors was 45.5 years, lower than the median age of 64 years for patients with monoclonal-negative specimens tested on the Optilite platform (n = 3217) as well as the median age encountered when testing for SFLC in clinical practice. Despite this caveat and the knowledge that KLR values increase modestly with age Figure 6A, results from these 106 normal blood donors nevertheless verified the new KLR interval of 0.67 to 2.13 and supported our in silico approach to defining a normal population.

DISCUSSION

Our systematic, informatics-based approach used data from 8178 clinical specimens to analyze SFLC assay performance in the context of contemporaneous immunofixation results generated over 1 decade at our institution. These analyses revealed that the manufacturer-defined normal reference intervals for the Freelite assay, used in our laboratory and specified in international guidelines, inappropriately categorized KLR values from individuals without monoclonal gammopathies as “abnormal.” We have therefore proposed new reference intervals to improve the accuracy of SFLC interpretation for the Optilite platform. We further identified a small proportion of LMFLC-containing specimens exhibiting normal KLRs, even after adjustment of the normal reference interval. Overall, we expect these findings to encourage a more cautious approach to interpreting discordant test results as well as a widespread reassessment of SFLC reference intervals and clinical guidelines.

Despite the increased sensitivity of the Freelite assay generally over immunofixation,^1,10 we found that LFLC and KLR results can be “normal” when small LMFLC bands are identified by immunofixation, consistent with a previous report.³⁸ Although the frequency of these SFLC/immunofixation-discordant cases was low (n = 30) and historical data were often limited, λ chain disease was confirmed by an orthogonal method in 8 of these individuals. Hypotheses that have previously been proposed to account for this discordance include the “amyloidogenic nature” of FLC,³⁹ preferential binding of Freelite detection antiserum to dimerized rather than monomeric LFLCs,⁴⁰ or issues of assay nonlinearity at low λ levels.^39,41-43 Testing with an alternative SFLC assay that uses different polyclonal antisera for detection did not resolve the λ discordance among the limited number of samples we tested. A newer method of identifying and quantifying monoclonal components by mass spectrometry has been described, but it, too, relies on polyclonal antisera for enrichment of immunoglobulins to achieve high sensitivity and does not display complete concordance with either immunofixation or the Freelite assay.⁴⁴ We therefore speculate that detection of particular LFLCs may be limited by the polyclonal antibodies present in different assay detection antisera, affecting the signal-to-noise ratio or detection of patient-specific LFLC epitopes against which it is difficult to raise antibodies. Although the cause of the λ discordance remains elusive, it is important to recognize that pathologic LMFLC bands can be identified by immunofixation when the SFLC assay yields “normal” results. Both laboratorians and clinicians should be aware of this possibility when selecting tests and interpreting discordant test results because missed diagnoses or treatment delays could result.

The major finding of this study is the need to modify Freelite SFLC assay reference intervals. Other investigators have described high Freelite KLR false-positive rates and right-shifted KLR distributions in both healthy controls and patient populations^37,38,45-48 and attributed these phenomena to technical issues such as assay drift and lot-to-lot calibration shifts.^37,46,48 Our comprehensive analysis shows that falsely high KLRs were produced from a monoclonal-negative population across multiple instrument platforms and could not be explained by patient age, sex, kidney impairment, or reagent lot changes over time. Although we also detected gradual assay drift and instrument bias over the course of our study, SFLC testing guidelines and volume also increased markedly during the study period, and all likely contributed to the upward shift in the KLR. Even at the earliest yearly time interval examined (2010), however, the median KLR of our monoclonal-negative samples was estimated to be 1.07 (among 102 of 5057 samples), higher than the originally reported median value of 0.6 from 2002.¹

Nonanalytical factors that could have contributed to the inappropriateness of the original manufacturer’s normal reference interval for our population include differences between the original Freelite reference population from Olmsted County and our regional population, such as patient age, sex, race, and health status. These differences could have been compounded by the substantial increase in testing volumes over the decade. Finally, our Optilite-only monoclonal-negative reference data set included more than 4-fold greater numbers of patients compared with the original reference interval study, which also could have affected the results. Although multiple factors could have driven the increase in KLRs, the critical issue for patient care is the need to address inappropriate reference intervals for our population (and those of other groups).

It is worth noting that approximately 66% of our population identified as non-Hispanic White, while 13.7% identified as Hispanic or Latino, 10% as Asian, 5.2% as 2 or more races, and 4.5% as Black or African American as of 2020 (among other demographic details outlined in the “Methods” section), which could affect the generalizability of our new SFLC reference intervals. Although the incidence of MGUS, multiple myeloma, and kidney failure have been found to be elevated in individuals with African ancestry,^49,50 our population has fewer Black or African American individuals than in the general US population, making it unlikely that this factor substantively affected the KLR distribution in our monoclonal-negative reference population. Furthermore, we purposefully chose not to analyze self-identified race/ethnicity data available in the electronic health record because these data are problematic. Collection of race/ethnicity data is not standardized, is inconsistently included in patient medical records, and can be markedly inaccurate.⁵¹ Race/ethnicity data are also imperfectly correlated to genetic ancestry information.⁵² Finally, our average annual incidence rate of multiple myeloma was slightly lower than the national average (6.6 vs 7 of 100,000), and our state ranked as the ninth healthiest overall among 50 US states, suggesting that our regional population is generally healthy.

Rightward shifts in the KLR reference interval have previously been attributed to the inclusion of individuals with impaired kidney function in healthy control populations.^32,33,36 In the recently published iStopMM study, more than 75,000 individuals from Iceland were screened for monoclonal gammopathy using the Freelite assay in addition to serum PE and immunofixation. The authors defined new “renal” KLR reference intervals from a population of more than 6000 individuals with different stages of kidney disease and no monoclonal gammopathy: 0.46 to 2.62 (eGFR = 45-59 mL/min/1.73m²), 0.48 to 3.38 (eGFR = 30-44 mL/min/1.73m²), and 0.54 to 3.30 (eGFR <30 mL/min/1.73m²).³⁶ Our Optilite-only data set included 108 patients with kidney disease only, identified by abnormal eGFR results and no other conditions known to increase SFLC (Supplemental Table 1); we judged these numbers to be too few to calculate reliable kidney disease reference intervals. Despite excluding patients with abnormal eGFR, however, our analysis demonstrated that a higher upper limit for the normal KLR reference interval is necessary for our population. We note that no KLR values were provided for the more than 45,000 iStopMM participants with SFLC data, normal kidney function, and no monoclonal gammopathy (also discussed by Minnema⁵³). Although the iStopMM data were obtained from a genetically homogenous Icelandic population, we believe that an analysis of this large data set could provide valuable information and potentially support the idea that the original Freelite reference interval warrants readjustment beyond our population.

The gold standard for verifying a normal reference interval for any assay performed in a clinical laboratory is to test samples from healthy donors. Indirect approaches for reference interval determination, as used in our study, are known to carry the risk of diseased patients being inadvertently included in the population used to define the reference interval.⁵⁴ We recognize that our analyses relied on available ICD codes and eGFR data, which are likely imperfect. The rate of abnormally high KLRs in our Optilite-only monoclonal-negative reference samples (12%), however, is similar to the false-positive rate previously published from 1182 healthy controls³⁷ as well as to the rate we observed from testing 106 healthy donor samples. Multiple key advantages of the indirect approach used here over traditional testing of healthy volunteers include the ability to use much larger numbers of specimens, the lack of a requirement for additional SFLC testing, and time and cost-effectiveness for the clinical laboratory. Furthermore, reference intervals established using mean and standard deviation require normality of the underlying data distribution. By contrast, our approach is nonparametric and broadly applicable to a wide range of clinical analytes. Perhaps most importantly, this approach enables the continuous monitoring of SFLC results from any such “reference” individuals to facilitate ongoing quality improvement, particularly in the setting of assay drift.

We demonstrated that adjusting KLR reference intervals to 0.67 to 2.13 reduced the false-positive rate among monoclonal-negative specimens and improved the accurate classification of LMFLC-containing specimens. For patients with κ-involved disease, patients with successfully treated disease and ratios above 1.65 but below 2.13 will no longer be classified as having persistently elevated KLRs. This finding may be particularly relevant to patients with multiple myeloma, for whom a decision about minimal residual disease testing on a bone marrow biopsy can be informed by the presence or absence of clinically detectable disease. In addition, increasing the lower limit of the KLR reference interval from 0.26 to 0.67 decreases the λ-to-κ excess required for an abnormal ratio from 3.85 to 1.49, which should improve detection of λ residual disease. We purposely chose to use the central 95% of SFLC values from our stringently defined, Optilite-tested reference population (n = 1199) for our normal reference interval. We did not select the wider diagnostic range to minimize the likelihood of missing low, abnormal KLR results in our diseased population (Supplementary Table 3).

We anticipate that changes in our reference intervals will improve assay interpretation and utility across the clinician spectrum. For primary care professionals screening patients for PCN, implementation of appropriate KLR reference intervals is likely to reduce additional, unnecessary testing of patients who would have been “flagged” as abnormal. Moreover, in an era when patients have immediate access to their test results, potential unnecessary patient stress and anxiety will be reduced. For specialists caring for patients with PCN, our revised KLR reference interval should improve identification of stringent complete remissions in multiple myeloma and AL amyloidosis,^55,56 thereby affecting prognosis and follow-up.

Our analyses indicate that the use of different platforms to test the same commercially available kit can affect results sufficiently to alter result interpretation by the laboratory. Other instrument platforms are currently being used with the Freelite assay and other assays for serum FLC testing are commercially available. This, coupled with the knowledge that interpretation of FLC values and the KLR often requires correlation of test results with clinical findings during disease treatment and remission, underscores the importance of accurately defining appropriate reference intervals for every platform and assay.

Reference interval evaluations and adjustments are routinely conducted in clinical laboratories and usually have a focused, local impact. This is unlikely to be the case for the Freelite assay. The reference interval changes we propose are based on the largest population analyzed to date on the Optilite platform and are consistent with those suggested by other groups; we agree that upper KLR thresholds should be higher than currently recommended by the manufacturer.^45,47 Together, these studies provide compelling evidence that clinical laboratories conducting this testing worldwide should reevaluate their SFLC reference intervals in an independent and unbiased manner. Changes to these reference intervals may affect treatment duration and intensity as well as whether patients are classified as complete responders. Furthermore, given that IMWG diagnostic and clinical care guidelines define “abnormal” using Freelite reference intervals established in 2002¹ and cited by the manufacturer, widespread adoption of new reference intervals will likely necessitate revision or amendment of these internationally used definitions and guidelines.

Data availability

The deidentified data sets analyzed in the current study are available from the corresponding author on reasonable request.

Conflict of interest disclosure

A.J.C. discloses the following: research funding from Harpoon, Sanofi, BMS, Janssen, Nektar, and AbbVie; consultancy for AbbVie, BMS, Allogene, EUSA, GlaxoSmithKline, Janssen, and Secura Bio; and advisory board membership at Adaptive Biotechnologies. D.G. discloses the following: research funding from Bristol Myers Squibb, Cellectar Biosciences, Janssen Biotech, Juno Therapeutics, Seattle Genetics, and SpringWorks Therapeutics; membership on a board or advisory committee at Bristol Myers Squibb, GlaxoSmithKline, Janssen Biotech, Neoleukin, and Seattle Genetics; patents or royalties from Bristol Myers Squibb and Juno Therapeutics; and consultancy for Legen Biotech. A.B. discloses the following: research funding from Shionogi and honoraria from the World Congress of Bronchology and Interventional Pulmonology. M.W. discloses the following: research funding from the National Institutes of Health; membership on a board or advisory committee at the College of American Pathologists (DIFCC), the American College of Rheumatology (CORC), and the International Union of Immunological Societies (Autoantibody Standardization Committee); patents or royalties from UpToDate and Medical Training Solutions; honoraria from the European Autoantibody Standardization Initiative; and ownership interests in UberDuck. C.M. discloses membership on a board or advisory committee at Merck.