Cut-off values of serum IgG4 among three reagents, including a novel IgG4 reagent: a multicenter study

Elevated serum IgG4 is a useful marker of IgG4-related disease (IgG4-RD) activity. However, there is no uniformity in the cut-off values of IgG4 among the various reagents. The aim of this study was to compare the measured and cut-off values of IgG4 assessed using three different reagents. This study enrolled 466 IgG4-RD and non-IgG4-RD patients who required measurement of serum IgG4 levels to diagnose or treat IgG4-RD. Serum IgG4 was measured using three reagents: N-assay LA IgG4 Nittobo (Nittobo), BS-NIA IgG4 (TBS), and N Latex IgG4 (Siemens). The values obtained using the three reagents were compared, and cut-off values were calculated for each. Although there was good correlation among the results with the three reagents, the measured and cut-off values were all different. The Nittobo values were 1.4 times the TBS values and the TBS values were almost half those of the Siemens values. ROC curve analysis showed cut-off values for the Nittobo, TBS, and Siemens reagents of 1.42, 1.31, and 2.38 g/L, respectively. The measured and cut-off values of serum IgG4 vary depending on the reagents used for the assay, although there is good correlation among the values measured by the three reagents.

www.nature.com/scientificreports/ addition, a new reagent has recently been released by Nittobo Medical Co., Ltd. (Nittobo, Tokyo, Japan) and has begun to be used in Japan. However, all three reagents result in different measured values 11 . This is reflected in the differences in their reference ranges: the upper limit of the reference range for the Siemens reagent is about twice that of the TBS reagent. A report on the cut-off value for IgG4-RD from China stated a value of 2.48 g/L, which is about twice the diagnostic standard of 1.35 g/L in Japan 12 . This is not due to differences in research subjects, but to differences in the reagents used.
Although there are many reports on calculation of the cut-off value for a single reagent, no reports have shown how the cut-off values differ between reagents. In this study, we report the results of simultaneous measurement of IgG4-RD in patient samples using reagents from the three companies, and calculation of the cut-off values for each reagent.

Results
Characteristics of patients and serum IgG4 levels. We grouped the 466 enrolled patients according to their diagnosis and treatment. All patients who had received some kind of treatment, including medications such as steroids and surgical treatment at the time of blood collection, were grouped as "during or posttreatment". The median serum IgG4 levels in each group are shown in Table 1. Three hundred forty-two patients (73%) were diagnosed with IgG4-RD (definite IgG4-RD group), 56 of whom were untreated at the time of blood collection and the remaining 286 of whom were evaluated during or after treatment. Thirty-six patients (8%) were suspected to have IgG4-RD but were not confirmed (suspicious IgG4-RD group), and the remaining 88 patients (19%) were diagnosed or suspected to have non-IgG4-RD. In the definite and suspicious IgG4-RD groups, 70% of patients were male, while 40% of patients in the non-IgG4-RD group were male. Median IgG4 levels in pre-treatment patients were significantly higher than those in during or post-treatment patients in the definite and suspicious IgG4-RD group, although there was no significant difference between the pretreatment and treatment subgroups in the non-IgG4-RD group. The same trend was observed for all three reagents. Comparing the median and interquartile range (IQR) of IgG4 levels measured with the three reagents, the values were in ascending order with TBS, Nittobo and Siemens reagents, except that values were similar with Nittobo and TBS reagents in the non-IgG4-RD group.
Correlations between the three reagents. Correlations between the three reagents (Nittobo, TBS and Siemens) for all 466 patients are shown in Fig. 1A,B,C. For all combinations of comparisons, the major-axis regression lines deviated from the diagonal line of equality (broken line). This indicated that the values measured by the three reagents were all different. However, when the analysis was limited to patients with IgG4 levels less than 1.5 g/L (Nittobo value) (Fig. 1D,E,F), the range included almost all healthy individuals, and the values obtained with Nittobo and TBS matched perfectly (Fig. 1D). In comparisons between Nittobo and Siemens, the slope of the regression line did not change in comparisons between all patients and for patients with IgG4 values below 1.5 g/L (Fig. 1B,E), with the values measured by Nittobo being almost 0.6 times lower than the value obtained with the Siemens reagent. In the comparison of IgG4 values measured by TBS and Siemens reagents, although the slope of the regression line was slightly different when comparing all patients versus patients with values below 1.5 g/L, the value with TBS was almost half that of the value with Siemens. www.nature.com/scientificreports/

Cut-off values for each reagent. ROC analysis was performed between untreated patients in the definite
IgG4-RD group (n = 56) and untreated and diagnosed with non-IgG4-RD patients in the non-IgG4-RD group (n = 41). Pre-treatment patients in the non-IgG4-RD group (n = 54) included both patients with confirmed non-IgG4-RD disease and those with suspected non-IgG4-RD disease. ROC analysis used data from only patients with confirmed non-IgG4-RD (n = 41). The involved organs in IgG4-RD patients and the diseases in non-IgG4-RD patients are shown in Table 2. Thirty-four patients in the IgG4-RD group had lesions in only a single organ, while 22 patients had lesions in more than one organ. Fifty-two patients (93%) had lesions in the lacrimal gland, salivary gland and/or pancreas. The most common disease in the non-IgG4-RD group was pancreatic cancer, followed by Sjögren syndrome and pancreatitis. The ROC curve, cut-off value and area under the curve (AUC), sensitivity and specificity for each reagent are shown in Fig. 2 and Table 3. The cut-off value of Nittobo was 1.42 g/L, which was close to the cut-off value of 1.31 g/L of TBS, although the cut-off value of Siemens was nearly double. Although the cut-off values differed among the three reagents, no significant differences in AUC, sensitivity, and specificity were observed.

Discussion
In this study, we evaluated correlations between three IgG4 reagents using 466 patients' samples, and calculated the cut-off value for each reagent. The correlation between the three reagents was compared over a wide range of values and patients, including not only IgG4-RD patients, but also patients with other diseases that need to be distinguished from IgG4-RD. The results showed that although there was good correlation between the results with the three reagents, the measured values were all different. As a result, the cut-off values also varied depending on the reagent used. This is the first study to simultaneously calculate the IgG4-RD cut-off value for the three IgG4 reagents. The cut-off value of 1.31 g/L of the TBS reagent in this study was almost the same as the value of 1.35 g/L determined using the TBS reagent in 2001 10 . The calculated cut-off value of 1.41 g/L for Nittobo was very close to that of TBS, and the slope of the regression equation of the correlation for values less than 1.5 g/L was 1.0. Therefore, a cut-off value of 1.35 g/L can be applied for the Nittobo reagent. On the other hand, a different cutoff value needs to be set for the Siemens reagent. This study calculated a cut-off value of 2.38 g/L for the Siemens Diagnosis of IgG4-RD might sometimes be difficult. This is because patients' characteristics differ depending on the affected organ, and because elevated serum IgG4 levels are not always observed in IgG4-RD patients 13 . Therefore, it is necessary to make a final diagnosis by a combination of evaluations, such as computed tomography, magnetic resonance imaging, and biopsy. However, as evidenced by the high AUC values in the ROC Table 2. Organs involved in the IgG4-RD group and diseases in the non-IgG4-RD group. ROC analysis was performed between 56 pre-treatment patients in the definite IgG4-RD group and 41 patients with confirmed non-IgG4-RD disease in pre-treatment patient of the non-IgG4-RD group.  www.nature.com/scientificreports/ analysis of this study, it is certain that many patients with IgG4-RD are identified by the presence of high serum IgG4 levels, and serum IgG4 levels contribute to the diagnosis of IgG4-RD. Moreover, it has the advantage of being easier to perform than other examinations. However, the differences in measured values between IgG4 assay kits can mislead clinicians when making a diagnosis. The cut-off value for serum IgG4 was described as 1.35 g/L in the Japanese comprehensive diagnostic criteria published in 2011, but this value could be used only for the TBS reagents used in Japan at the time. American College of Rheumatology/European League Against Rheumatism classification criteria for IgG4-RD adopted the upper limit of the reference range of each assay kit as the cut-off level for diagnosis 14 . Since several reagents are used in the United States, it is considered that the reference value of each reagent should be used instead of a specific number. Other measurement kits might become available in different countries in the future. To avoid confusion, harmonization of the measurement values between the assay kits should be achieved. Differences in IgG subclass values between TBS and Siemens reagents have been previously reported 15 . Although harmonization of the value of total IgG has been achieved using Certified Reference Material (CRM) 470 released by the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) in 1993, CRM 470 has no certified values for IgG subclasses and the measurements have not been unified. TBS uses CRM 470 16 and Siemens uses WHO67/97 as the reference material 17 , IgG4 values of each reference material were determined by different researchers. Nittobo has the original master calibrator with IgG4 value determined themselves, and they trace the value to the kit's calibrator 11 . Those mean that all three companies have established the different traceability. Consequently, the difference in the slope of the correlation results below 1.5 g/L between Siemens and the other two kits could be induced by the difference in the determined values for their own calibrators. According to the report of Ladwig et al., the quantitative value of IgG subclasses measured using LC-MS/ MS are consistent with the values measured using the TBS reagent 18 . Bernasconi et al. reported that the amount of purified IgG4 protein was consistent with the value of IgG4 measured using the TBS reagent 19 . These reports suggest that the IgG4 value measured using the TBS reagent is an accurate value.
However, the different calibrators, assessed by the original traceability of each company, cannot explain the mismatch in the slopes of the regression equations between the correlation diagrams for all data and those for the data less than 1.5 g/L in the comparisons between TBS and other two methods. This would mean that even if all methods indicate sufficient linearity in the localized range such as less than 1.5 g/L, no absolute linearity in wide range, such as from 0 to 20 ~ 40 g/L, is provided by all or some of those methods although apparently sufficient linearity is observed in three methods. Therefore, a data around cut off value, which is practically important to diagnose IgG4-RD, is convertible using a constant factor among three methods. However, although the good correlation coefficient for all data was also observed among three methods, it does not mean to be able to convert a data obtained by each method using the same factor described above, especially in a data much higher than cut off value. Imoto et al. reported that the sum of the values of the four IgG subclasses differs from the measured value of IgG when IgG4 levels increase 20 . In healthy individuals with a normal balance of the four subclasses, the total IgG value and sum of the subclass values that is measured using the TBS measurement kit are the same, but divergence occurs when IgG4 is high. From our result, it is considered that the divergence reported by Imoto et al. is due to poor linearity in the high value region in IgG4 measurement. Further validation of the accuracy of the measurement value with elevated IgG4 levels will be needed.
In conclusion, serum IgG4 is a useful marker for IgG4-RD diagnosis and treatment, but the values vary greatly depending on the reagents used for the assay. Therefore, it is necessary to change the cut-off value according to the reagent used. In previous studies, the cut-off value was calculated for only a single reagent, but in this study, cut-off values for three reagents were calculated simultaneously. Any reagents described here are available to diagnose IgG4-RD if the suitable cut-off value for each reagent is used. However, it is disclosed that the proportional conversion for all of the data is not always possible. Therefore, when using serum IgG4 values for the diagnosis and treatment of IgG4-RD, it is essential to confirm which reagent is used. It is hoped that harmonization of IgG4 measurements will be achieved in the near future.

Methods
Subjects. This prospective study included not only patients with IgG4-related disease, but also those with differential diseases. Furthermore, we targeted patients who required serum IgG4 measurement for diagnosis or treatment regardless of the presence or absence of a definitive diagnosis and the stage of treatment. We enrolled 466 patients from Sep 2016 to Jan 2017 at 23 hospitals in Japan. The diagnosis of IgG4-RD was made according to the comprehensive diagnostic criteria for IgG4-RD, 2011 (21). The study was conducted according to the guidelines of the Declaration of Helsinki and was approved by the institutional review board of Shinshu University School of Medicine (Matsumoto, Japan) (approval no. 3325). Written informed consent was obtained from all subjects.