To the Editor:
IgG4-related disease (IgG4-RD) is a chronic inflammatory-fibrosing disorder affecting
virtually any organ, but most frequently the pancreaticobiliary system. Differential
diagnosis is challenging and relies on multiple criteria rather than on single markers.
Serum IgG4-levels (sIgG4) are elevated in IgG4-RD, correlate with disease activity,
yet lack sensitivity. Malignancies can also display elevated sIgG4 levels that reduce
The correct differential diagnosis of IgG4-RD from malignancies is critical to implement
treatment and avoid unnecessary resections. Thus, reliable biomarkers are urgently
Previously, qPCR-based measurements of the IgG4/IgG mRNA ratio that originate from
dominant IgG4+-B-cell receptor (BCR) clones have been proposed to fill this gap with
“ideal” test characteristics (AUC 0.991, sensitivity 94%, specificity 98.7%).
In contrast, recent work from Beuers and colleagues found relevant limitations of
IgG4/IgG qPCR ratio in distinguishing between IgG4-RD and pancreaticobiliary malignancies.
Our data strongly support this observation from an independent referral center in
distinct cohorts comprising the diagnostic spectrum of pancreaticobiliary IgG4-RD:
Specifically, we analyzed IgG4/IgG qPCR ratios in blood samples from our prospectively
collecting biobank (ethics vote No. 159/19 and 87/20) in a cohort of 98 patients with
IgG4-RD, cholangiocarcinomas, pancreatic ductal adenocarcinomas and benign pancreaticobiliary
diseases such as chronic pancreatitis (CP) and primary sclerosing cholangitis (PSC)
(Fig. 1A). mRNA was extracted from leucocytes, followed by qPCR analyses and IgG4/IgG
ratio calculation (cut-off range from 3.5–6.0%) (Fig. 1B
). Remarkably, false positive test results became particularly evident in patients
with cholangiocarcinomas (12–18 out of 25; 48–72%) and pancreatic carcinomas (9–10
out of 21; 42.9–47.6%). However, false-positive diagnoses were also made in CP and
PSC groups (Fig. 1B,H). In contrast, 10-13 out of 16 patients with confirmed IgG4-RD
could be correctly diagnosed leading to a sensitivity of 62.5–81.3% depending on the
3.5 or 6.0% cut-off, respectively (Fig. 1B). However, our IgG4-RD cohort was not therapy-naïve,
lowering sensitivity as previously reported.
To circumvent this problem, we challenged our non-IgG4-RD patients with the treatment
naïve IgG4-RD cohort from Doorenspleet et al. during area under the receiver-operating
characteristic curve (AUC) analysis. Still, AUC dropped from 0.987 to 0.863 (Fig. 1C),
pointing to a lower test performance, independent of the treatment status of IgG4-RD
patients. There might be several reasons for this observation: (i) Most patients in
de Vries et al.
or in our study have not received chemotherapy, in contrast to the original publication.
Hence, chemotherapy might also suppress existing BCR-clones in malignancies, therefore,
limiting diagnostic accuracy, as shown for BCR-counterparts in IgG4-RD.
(ii) We hypothesized that technical and methodological constraints in the qPCR-reaction
might hamper test performance. In particular, a common single nucleotide polymorphism
(RS10137020; Fig. 1D), with a minor allele frequency of 30% in one of the primer binding
sites, could reduce the efficacy of PCR. Furthermore, separated reactions and primer
sets for IgG4 and total IgG mRNA measurement, declining polymerase activity, PCR efficacy
and pipetting errors could negatively influence the test performance. To circumvent
this, we designed a multiplex digital droplet PCR assay simultaneously measuring IgG4
and total IgG: Specifically, the IgG4/IgG ratio was calculated based on Poisson distribution
using a single pair of primers implemented together with allele-specific TaqMan®-probes
(Fig. 1E; IgG4/IgG digital droplet PCR). Intriguingly, the digital droplet PCR approach
did not outperform the standard qPCR-assay
for determining IgG4/IgG ratios. This excludes technical constraints in the method
but instead suggests reduced test performance (Fig. 1F). To further probe IgG4/IgG
qPCR ratio against the current gold standard in diagnostic scorings, AUC analysis
for pre-therapeutic sIgG4 levels was performed in our cohorts and compared to Doorenspleet
Conversely, AUC analysis for sIgG4 levels across various comparisons displayed similar
accuracy values, underpinning the superiority of sIgG4 levels compared to IgG4/IgG
ratios in the differential diagnosis of IgG4-RD (Fig. 1C,F,G;AUC 0.902 vs. 0.916 vs.
0.950). False positive diagnosis of IgG4-RD in case of pancreaticobiliary malignancy
can delay diagnosis, potentially curative surgery or timely chemotherapy. Also, in
case of benign disease unnecessary steroid pulse treatment can cause side effects.
Therefore, false positive test results pose the strongest threat to non-IgG4-RD patients
and need to be imperatively avoided, likewise underpinned by de Vires et al.
In turn, we compared false positive rates for IgG4/IgG mRNA ratios with sIgG4 levels
head-to-head for differential diagnosis of (non-)IgG4-RD. Strikingly, there were lower
false positive rates for sIgG4 levels compared to IgG4/IgG mRNA ratios in both patients
with malignant and benign non-IgG4-RD (Fig. 1H). Collectively, we demonstrated (i)
in concordance with de Vires et al. (ii) on independent cohorts and (iii) with different
methodological setups that the IgG4/IgG mRNA ratio is prone to false-positive results,
which could cause misdiagnosis of pancreaticobiliary cancer. Furthermore, we found
that sIgG4 levels were more accurate, although still not perfect. Therefore, our study
questions the clinical benefit of IgG4/IgG mRNA ratios in the differential diagnosis
of IgG4-RD in support of de Vires et al. but concurrently underpins the necessity
for more reliable biomarkers to differentially diagnose IgG4-RD.
IgG4/IgG mRNA ratios are not suitable for differential diagnosis of IgG4-RD.
(A) Patient characteristics of internal study cohort: total cohort, IgG4-RD and non-IgG4-RD
subgroups. For non-IgG4-RD patients, data on former glucocorticoid treatment was not
available. Immunosuppression refers to 2 cases of rituximab and 1 case of azathioprine
treatment. Infections were 1 case each of infected walled-off necrosis (IgG4-RD patient),
bronchitis, acute flare of ulcerative colitis and post-ERCP pancreatitis (each in
non-IgG4-RD patients) (B) IgG4/IgG mRNA expression in patients with IgG4-RD or non-IgG4-RD
differential diagnosis, measured by qPCR. Dashed lines: Cut-off values (6.0% and 3.5%
relative expression of IgG4). Green, orange or red dots: Value below, between or above
cut-off values, respectively. (C) AUC of IgG4/IgG mRNA ratio. Green line: Test results
as published by Doorenspleet et al. (IgG4-RD vs. non-IgG4-RD; external data
); black line: IgG4/IgG mRNA ratio of IgG4-RD patients from
compared to non-IgG4-RD patients from own cohort. (D) Schematic illustration of the
SNP RS10137020 with a minor allelic frequency of ∼30%. (E) Principle of IgG4/IgG based
digital droplet PCR approach. Concentration was calculated based on Poisson distribution
(BioRad® QuantaSoft™ 1.7.4.0917) (F) AUC of IgG4/IgG mRNA ratio either assessed by
quantitative PCR (quant. PCR, blue line) or by digital droplet PCR (dig. PCR, red
line). (G) AUC analysis of sIgG4 levels in distinct cohorts. Black line: Own cohort
of IgG4-RD vs. non-IgG4-RD patients. Green line: Data from Doorenspleet et al. on
sIgG4 levels in IgG4-RD vs. non-IgG4-RD;
Blue line: sIgG4 levels of IgG4-RD patients from
compared to non-IgG4-RD patients from own cohort in analogy to (C). (H) False positive
rates of IgG4/IgG mRNA ratio (qPCR) and serum IgG4 levels (sIgG4; cut-off: 135 mg/dl)
according to published cut-off values (patients included when both sIgG and IgG4/IgG
mRNA qPCR were available, n = 69). AIP, autoimmune pancreatitis; AUC, area under the
receiver-operating characteristic curve (R 3.6.0, pROC package); CCC, cholangiocarcinoma;
CP, chronic pancreatitis of other causes; ERCP, endoscopic retrograde cholangiopancreatography;
IgG4-RD, IgG4-related disease; PDAC, pancreatic ductal adenocarcinoma; PSC, primary
sclerosing cholangitis; SNP, Single nucleotide polymorphism.
Main funding is provided by the
(DFG) K.L. 2544/7-1, 1-1, 1-2 and 5-1 and the “Heisenberg-Programm” KL 2544/6-1. A.K.
and T.S. are PIs in the HEIST RTG funded by the
GRK 2254/1. F.A. is a HEIST fellow. A.K. is a fellow of Else-Kröner-Fresenius Excellence
program. L.S. is a fellow of the Clinicain Scientist Programm of Ulm University. L.P.
receives funding form Bausteinprogramm of Ulm University.
L.S.: Experiments, data evaluation, text drafting, figures, clinical data collection.
F.A.: Experiments, data evaluation, figures. F.S.: Experiments. J.B., L.P., A.B, G.B.,
F.U.W.: Clinical data collection. T.S. text editing, approval. M.M., A.K.: Supervision,
idea, data evaluation, text drafting, figures, clinical data collection.
Conflict of interest
The authors declare no conflicts of interest that pertain to this work.
Please refer to the accompanying ICMJE disclosure forms for further details.