Toward standardization of carbohydrate-deficient transferrin (CDT) measurements: III. Performance of native serum and serum spiked with disialotransferrin proves that harmonization of CDT assays is possible

Introduction

Carbohydrate-deficient transferrin (CDT) is a generic term that refers to the transferrin glycoforms whose concentration in blood is temporarily increased by sustained alcohol consumption [1]. Transferrin consists of a single polypeptide chain with 679 amino acids with two N-linked complex oligosaccharide chains. The microheterogeneity in the oligosaccharide chains is important: each of the two chains can be bi-, tri- or rarely tetra-antennary; each antenna usually bears a sialic acid residue at its end and the glycoform is named by the total number of sialic acid residues [2]. The main component of serum transferrin, tetrasialotransferrin, consists of two biantennary chains and makes up approximately 80% of the total. Other glycoforms usually present in normal samples are pentasialotransferrin (approx. 15%), trisialotransferrin (approx. 4%–5%), disialotransferrin (approx. 1.0%–1.5%) and hexasialotransferrin (approx. 1%) [3, 4].

An alcohol consumption averaging at least 50–60 g/day for more than 2 weeks is considered necessary to induce a relative increase of disialotransferrin. When the disialotransferrin level reaches approximately two- to three-fold the upper limit of the reference values, asialotransferrin usually becomes measurable [5]. Due to the high specificity for alcohol increasing disialotransferrin, CDT was proposed as a biomarker of heavy alcohol use and has been widely used as such. Over the last 20 years, a number of CDT methods were developed based on different analytical techniques and principles, however, without any standardization of the reference intervals and cut-offs. Unfortunately, a clear unifying definition of the measurand (‘CDT’) has been hard to give, due to the different principles of methods employed. Initially, the sum of all glycoforms with a pI ≥5.7, corresponding to disialotransferrin and the rare mono- and asialotransferrin, was referred to as CDT [1]. Some older methods also included part of trisialotransferrin. More recently, a direct immunoassay based on a monoclonal antibody that recognizes the transferrin glycoforms lacking at least one of the glycan chains was developed [6]. In the latter case, the measurand is probably a mixture of disialo-, monosialo- and asialotransferrin.

As a consequence, neither the reference limits nor the cut-off values were similar across assay methods, which hampered understanding of the diagnostic value of CDT and its routine use. This prompted the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) to initiate a Working Group aiming for standardization of CDT measurements (WG-CDT). In its first publication [7], disialotransferrin (disialylated monoglycan transferrin) was defined as the measurand for a reference method and the target analyte for the standardization work, and high-performance liquid chromatography (HPLC) [8] with photometric detection was proposed as the candidate reference method. A network of reference laboratories running this method was formed demonstrating good intra- and inter-laboratory performance [9].

Recently, the WG-CDT focused their efforts on the development of secondary reference materials to be used by the manufacturers to standardize their test results to the reference method. These materials must be commutable, to allow harmonization of all different methods currently in use, and be stable in storage.

The present study investigated the suitability of native and disialotransferrin-spiked serum panels as candidate secondary CDT reference material. This was done by testing the commutability in the reference laboratories, the manufacturers’ laboratories, and in two external quality assurance (EQA) schemes. Whether a single or multiple point calibration is needed for calibration of the present commercial methods was also examined.

Samples

The study was based on the use of two six-level series of samples with similar disialotransferrin concentration ranges, one being constituted from native human serum samples and the other formed by spiking a low CDT serum sample with isolated disialotransferrin.

Initially, two serum pools were prepared from anonymous leftover volumes of samples from routine examinations in abstaining patients (pool P1) and alcoholic patients (pool P6). The CDT level in these pools was determined using the HPLC candidate reference method [8] and the relative concentrations of disialotransferrin were found to be approximately 1.2% for P1 and approximately 8.0% for P6. Then, four intermediate CDT levels (P2–P5) were produced by a mathematical dilution (equal intervals) of P1 and P6 to obtain disialotransferrin concentrations of 2.6%, 3.9%, 5.3% and 6.6%. These six levels constituted the patient (P) panel.

A second panel was prepared by spiking another aliquot of the P1 pool with purified disialotransferrin, obtained by courtesy of Bio-Rad (Munich, Germany) [10], to a final disialotransferrin concentration of 8.8% (pool S6). As for the patient panel, four intermediate CDT levels (S2–S5) were prepared by dilution of P1 and S6 to obtain disialotransferrin concentrations of 2.7%, 4.2%, 5.8% and 7.3%. These six levels constituted the spiked (S) panel.

All samples were distributed into aliquots and lyophilized using the procedure described in a previous study [9]. It was found in preliminary experiments that these samples were stable for at least 2 years when stored frozen at –20°C (Weykamp and Wielders, unpublished).

Analytical methods

The commutability of the patient and spiked panels was assessed with the most frequently used commercial CDT methods, which are based on various analytical principles. Three capillary zone electrophoresis (CE) methods were included: the CEofix CDT procedure [11] from Analis (Suarlée, Belgium) and the Capillarys [12] and Minicap methods of Sebia (Lisses, France). Four HPLC methods were included: the candidate reference method [8], the %CDT by HPLC assay of Bio-Rad (Munich, Germany) with both standard [13] and fast procedures, and the ClinRep CDT [14] of Recipe (Munich, Germany). The immunonephelometric N Latex CDT assay [6] of Siemens (Marburg, Germany) was also included.

All CDT values by HPLC and CE methods in this study are expressed as disialotransferrin percentage of total transferrin (%DST). With HPLC, %DST represents the area under the disialotransferrin peak compared to the total peak area of all transferrin glycoforms as monitored at 470 nm. This wavelength is specific for the iron-transferrin complex. The same glycoform pattern is monitored at 200–214 nm using the CE techniques, assuming no interferences from other proteins with similar electrophoretic mobility. For the N Latex CDT immunoassay (measuring transferrin glycoforms lacking one or both glycan chains), values are expressed as percentage of total transferrin, which is measured in a second separate assay.

Analytical performance with the HPLC candidate reference method

Analytical performance with the commercial CDT methods

Origin of non-commutability

There are three major effects causing non-commutability of analytical results that are attributable to the sample: 1) the matrix (biological or artificial) in which the measurand is dissolved may interfere with the analytical method; 2) pretreatment (e.g., lyophilization) of the sample may cause changes in physical, chemical or immunochemical (i.e., epitope recognition) properties of the measurand; and 3) the nature of the measurand in that a native protein may not react exactly as a synthetic protein or an isolated and spiked protein, leading to a variation in the analytical result for native samples when a synthetic/isolated protein is used as calibrator. In a previous WG-CDT study [9], matrix effects due to 1) and 2) were excluded; the present study therefore focused on the nature of the measurand.

Figure 1

%CDT (N Latex CDT immunoassay) and %disialotransferrin (HPLC and CE methods) results for patient and spiked panels using commercial CDT methods plotted against the %disialotransferrin (%DST) values assigned by the network of reference laboratories using the HPLC candidate reference method.

(A) Patient serum panel (P1–P6). (B) Disialotransferrin-spiked serum panel (S1–S6).

When regression (measured value vs. theoretical value) equations slopes and/or the intercept to the origin differ between native materials and synthetic materials, this can be attributed to the nature (native or synthetic) of the measurand. We calculated the difference (patient panel value – spiked panel value) using the regression equations corresponding to the data respectively presented Figure 1A and 1B for all methods. These differences are graphically presented in Figure 2. The three HPLC methods show very small differences across the range 1%–5%, while other methods are characterized by positive or negative slopes. We quantified the differences at two CDT concentrations (1.5% and 4%). At 1.5% CDT level, the differences (patient panel – spiked panel) ranged from −0.11% (immunoassay, N Latex CDT) to 0.36% (CE, Sebia Minicap). At 4%, the between analyte differences were comprised between −0.71% (CE, Sebia Minicap) and 0.41% (CE, Analis). These differences in values between the spiked and patient samples were considered as reflecting the non-commutability attributed to the nature of the measurand (Figure 2). This graph indicated that both panels are commutable for the HPLC methods, but not for the CE and immunonephelometric methods. Thus, calibration of CE and immunonephelometric methods is not possible with certified reference materials (CRM) spiked with DST; hence, the only option to prepare CRM suitable for all method principles is to manufacture them from patient samples. As these results were obtained in panels prepared by pooling several samples, a similar study will be repeated using individual patient samples following the CLSI C53A guideline.

Calibration potential in EQA schemes

The design of the study allowed us to simulate calibration and thus to establish the harmonization potential. Using the six-level patient and spiked panels, the manufacturers established correlation equations for their CDT methods against the HPLC candidate reference method. These equations were used to recalculate the mean values from the participants of the Equalis (Sweden; n=50) and the SKML (The Netherlands; n=20) external quality assurance (EQA) programs for CDT. In this experiment, the results obtained by the laboratories of the manufacturers and the participants in the EQA schemes were merged both for the Bio-Rad HPLC (standard and fast versions) and Sebia CE (Capillarys and Minicap versions) methods. For the other commercial methods, the number of participants was too low to be useful for this approach.

Figure 2

Non-commutability of disialotransferrin-spiked serum calibrators with the commercial CDT methods at %disialotransferrin (%DST) concentrations ranging up to 5.0%, as assigned by the network of reference laboratories using the HPLC candidate reference method.

The difference between results with native samples harmonization and spiked samples harmonization is given on the y-axis.

Three of the patient and spiked samples were used (P1, P3 and S3). Table 2 shows the high between-method CV before calculation and its substantial reduction after the introduction of the corrective equation in both native samples. On the contrary, the between-method imprecision remained high and unchanged when using the calibration regression equation provided by the spiked panel. At low levels, the between-method CV was reduced from 40% to 6% in the patient panel. At high CDT levels, the intermediate CV was reduced from 7% to 3% in the patient panel but remained high (10%–12%) in the spiked panel. This confirms that CDT harmonization can be achieved by using a multi-level panel of secondary calibrators of native (human serum) origin, but not with disialotransferrin-spiked samples.

Conclusions

These data demonstrate an excellent linearity and reproducibility of results for the HPLC candidate reference method employed by the reference laboratories. Multi-level calibration using candidate secondary calibrators marginally improved the already high inter-laboratory agreement. The commercial CDT methods gave linear results over the studied concentration range with the candidate secondary calibrators. Harmonization of results can therefore be obtained with the use of calibration by a multi-level panel of calibrators based on human serum. These promising results were observed using sample pools and have to be confirmed in a further experiment based on samples from individual patients, according to the CLSI C53A guideline. The next and final step of the WG-CDT will be a full standardization of commercial CDT methods against a reference method, using certified reference materials.