Introduction
Autoimmune pancreatitis (AIP) is a rare form of chronic pancreatitis that was first proposed by Yoshida et al. [
1] in 1995. The clinical profile of AIP includes enlargement of the pancreas, irregular narrowing of the pancreatic duct, and the involvement of multiple organs [
2]. Type 1 AIP is an immunoglobulin (Ig) G4-related disease with increased serum IgG
4 levels and a good response to corticosteroids. It is a benign relapsing disease with a treatment rate of 97.9%; however, its overall relapse rate ranges from 20 to 60% [
3‐
6]. Furthermore, relapse seriously affects the lives of patients and comprises a heavy psychological burden; therefore, it is necessary to identify predictors of relapse during clinical diagnosis and treatment [
6].
Unfortunately, there is no definite consensus regarding the predictive factors for AIP relapse. Some studies have reported that initially high serum IgG
4 levels, persistently high serum IgG
4 levels after treatment, diffuse enlargement of the pancreas, proximal IgG
4 sclerosing cholangitis, and extensive multiorgan involvement may be risk factors of relapse [
5,
7‐
10]. Recently,
18F-fluorodeoxyglucose positron emission tomography/computed tomography (
18F-FDG PET/CT) metabolic parameters have been found to have good clinical value for determining the prognosis of many diseases [
11,
12].
18F-FDG PET/CT is advantageous when used for determining the differential diagnosis, performing early evaluation of the involved organ location, evaluating the lesion severity, and monitoring the response to AIP treatment [
13]. However, the predictive value of
18F-FDG PET/CT for AIP relapse is unknown. This study is the first to explore the predictive value of
18F-FDG PET/CT metabolic parameters to identify patients at risk for relapse before treatment.
Discussion
To the best of our knowledge, this is the first study to explore 18F-FDG PET/CT metabolic parameters to predict type 1 AIP relapse. We found that pretreatment 18F-FDG PET/CT metabolic parameters may be useful predictors of relapse, and that TLG had better ability to predict type 1 AIP relapse when the SUV fixed threshold was 2.5. We also established a multiparameter model based on IgG4 and TLG2.5 with an estimated predictive performance of AUC 0.806 that could significantly improve the ability to predict type 1 AIP relapse.
Type 1 AIP, also known as lymphoplasmacytic sclerosing pancreatitis, is the manifestation of pancreatic involvement in IgG
4-related diseases [
14,
20]. It is characterized by a high relapse rate ranging from 20 to 60% [
3‐
6]. Our cohort showed that the relapse rates of patients who did not receive maintenance therapy with low doses of glucocorticoids were as high as 64.7%. AIP relapse increases the risk of pancreatic atrophy, resulting in significantly decreased pancreatic function [
3,
5]. Although some risk factors for relapse have been proposed [
5,
7‐
10], the predictors of relapse remain poorly understood [
10].
The serum IgG
4 level is an important serological marker in clinical practice. Several studies have found that increased serum IgG
4 levels at diagnosis may predict relapse [
9,
21,
22]. However, the relationship between IgG
4 levels and relapse has been debated from different perspectives, and some studies have reported that these are not linked [
6]. Our data demonstrate that serum IgG
4 levels in the relapsed group were significantly higher than that in the non-relapsed group, and the optimal cutoff value was 946.5 mg/dL (approximately five times the ULN; OR, 1.001; 95% CI, 1.000–1.002). This is similar to the recommendation of the international consensus for the treatment of AIP, which suggests that a markedly increased serum IgG
4 level (such as > 4-times the ULN) before treatment is a predictor of relapse [
10]. Our results show that serum IgG
4 levels had relatively high predictive specificity (0.778); however, the predictive sensitivity (0.667) of serum IgG
4 alone is relatively low.
18F-FDG PET/CT is a noninvasive diagnostic tool that can provide anatomical morphology and metabolic information of the whole body with a single scan [
18]. The prognostic value of
18F-FDG PET/CT metabolic parameters has been demonstrated for many diseases [
11,
12,
23]. However, there are no reports of the prognosis determined using PET/CT for patients with AIP. We sought to explore whether
18F-FDG PET/CT could predict the recurrence of type 1 AIP. We found that, without low-dose glucocorticoid maintenance therapy, the
18F-FDG PET/CT metabolic parameters SUV
max, SUVR, and TLG
2.5 of the relapsed group were higher than those of the non-relapsed group in univariable analysis, suggesting that they may be potential risk factors for recurrence. Furthermore, based on multivariable analysis, we found that TLG
2.5 (OR, 1.001; 95% CI, 1.002–1.013; AUC = 0.678) may be a better predictor of relapse among all
18F-FDG metabolic parameters. Type 1 AIP is an immune-mediated inflammatory disease involving lymphoplasmacytes that are rich in IgG
4 infiltrates and fibrosis of the tissue [
20]. Because of the increased
18F-FDG accumulation in inflammatory lesions,
18F-FDG PET/CT may contribute to the presence of pancreatic and extrapancreatic organs in AIP [
24,
25]. TLG is a semi-quantitative parameter that combines metabolism and volume and can reflect the activity of glucose metabolism in the overall lesion [
26]. Our results showed that patients with higher TLG are more likely to experience recurrence. TLG, as a volumetric parameter of
18F-FDG PET, may reflect a wider accumulation of IgG
4 in the pancreas, greater glucose consumption in the lesions, and stronger activity. Therefore, these patients may require longer treatment and more effective medications to prevent recurrence. TLG
2.5 had better predictive specificity (0.778); however, the predictive sensitivity of TLG
2.5 (0.606) was relatively poor according to our results. To enhance the predictive performance, we identified two independent predictors, IgG
4 and TLG
2.5, that were associated with relapse, and developed a multiparameter model incorporating both factors. The use of multiple parameters in the model served to overcome the limitations of relying on a single parameter alone. The Hosmer-Lemeshow test and internal validation results indicate that the model has reasonable calibration. To facilitate clinical application, we have created a nomogram that predicts the likelihood of AIP recurrence based on TLG
2.5 and IgG
4 values.
We analyzed pancreatic and extrapancreatic organs by qualitative method to show other findings. Our data showed that the proportion of diffuse/multifocal pancreatic uptake was significantly higher in the relapsed group than in the non-relapsed group. Both qualitative and semi-quantitative (TLG
2.5) results also suggested that the larger the volume of pancreatic involvement, the more extensive the inflammatory activity, and that the higher metabolic activity of the lesion may lead to a higher risk of AIP recurrence. This result is similar to the previous treatment consensus [
10]. Although it was similar between groups, a nonstatistical difference in proportion of salivary gland involvement was higher in the relapsed group when compared with the non-relapsed group (10 [30.3%] vs. 2 [11.1%],
P = 0.231). These results were similar to those of studies by Yasutaka and Kuruma et al., who found that type 1 AIP accompanied by salivary gland involvement predicted relapse [
27,
28]. Previous studies have found that AIP with biliary tract involvement had a greater tendency to relapse; however, our results show that there was no significant difference in the proportion of biliary tract involvement between the two groups, which is contrary to our initial perception. The possible reasons for this discrepancy could be attributed to the poorer visualization of the bile ducts with
18F-FDG PET/CT as compared to MRI biliary imaging due to its lower sensitivity, along with possible factors such as sample bias and insufficient sample size. Even though our research suggests that the number of extrapancreatic organs with
18F-FDG uptake and affected lymph nodes before treatment were not distinctly different between the relapsed and non-relapsed groups, we must note that the distribution of systemic organs was sufficiently assessed using
18F-FDG PET/CT within our study population. These results are helpful in determining the degree of disease activity and differential diagnosis.
This study had some limitations. First, although this is the first study to explore 18F-FDG PET/CT metabolic parameters for predicting type 1 AIP relapse, the retrospective cohort study involving different treatment protocols may have a confounding effect on prognostication. In addition, the small sample size may limit the study results. Therefore, it is necessary to conduct large-scale, prospective validation studies among a homogeneous population of patients with AIP. Second, our study lacked external validation. Given the rarity of AIP and limited sample size, we were only able to perform internal validation, which nevertheless yielded promising results. We plan to gradually incorporate additional cases for external validation in the future. Third, although our study included patients from two medical centers and ensured similarity in the basic characteristics between treatment groups, it should be noted that the SUV is influenced by various factors, which may result in lack of repeatability of the model developed with metabolic parameters in clinical practice. In order to enhance the stability and repeatability of the model, we plan to incorporate additional PET image omics parameters, such as texture parameters, in our future studies.
In conclusion, a single PET/CT scan prior to treatment can aid in both diagnosing and assessing the distribution of lesions throughout the body, as well as provide a determination of the probability of recurrence. Among the 18F-FDG PET metabolic parameters, TLG2.5 is a better predictor of type 1 AIP relapse. Multiparameter models based on IgG4 and TLG2.5 can help to predict type 1 AIP relapse, and to some extent, screen patients who are more likely to benefit from maintenance therapy.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.