Baseline characteristics of IgG4-RD patients in the cohort
A total of 602 patients with IgG4-RD meeting the inclusion criteria in Figure 1 were enrolled in this study. The baseline characteristics are presented in Table 1. According to the 2019 ACR/EULAR IgG4-RD classification criteria , 59.63% patients in our cohort were male. The mean age at IgG4-RD diagnosis was 54.6±13.4 years. The median follow-up time of the patients was 47.0 (27.0-65.0) months. The most frequently involved organs were the submandibular glands (56.81%), followed by the lymph nodes (48.01%) and the lacrimal glands (41.36%). 29 patients (18 males and 11 females) were identified as IgG4-RD accompanied by malignancy. Of all cases with malignancy, the mean age was 58.9±12.1 years at IgG4-RD diagnosis and 56.8±13.7 years at malignancy diagnosis, respetively.
Clinical characteristics of IgG4-RD patients with malignancies
Baseline clinical characteristics in patients with and without malignancies were compared as shown in Table 1. There were no significant differences in demographic data, personal history and past medical history between IgG4-RD patients with and without malignancies. However, we observed statistical differences in laboratory results. The laboratory results revealed that IgG4-RD patients with malignancies had lower serum AGR (1.15 vs. 1.49, P＜0.001), higher serum IgG level (21.94 vs. 17.36 g/L, P=0.015) and higher eosinophil percentage (6.10% vs. 2.30%, P＜0.001) than those without malignancies. In terms of the distribution of organ involvement, submandibular glands, lymph nodes and lacrimal glands were the most frequently affected organs whether the patients developed malignancies or not.
Types of malignancy in Chinese IgG4-RD patients and SPRs
Clinical characteristics of the IgG4-RD patients with malignancies are shown in Table S1. 14 (48.3%), 3 (10.3%) and 12 (41.4%) patients developed malignancies before, on and after the diagnosis of IgG4-RD, respectively. 25 out of 29 (86.2%) IgG4-RD patients were diagnosed with solid tumours, which consisted of lung cancer, stomach cancer, cervical cancer, thyroid cancer, bladder cancer, testicular cnacer, kidney cancer, intra-abdominal soft tissue sarcoma, colon cancer, prostate cancer, pancreatic cancer and esophageal squamous cell carcinoma (ESCC). Another 4 (13.8%) patients developed haematological malignancy, including 2 non-Hodgkin’s lymphoma (NHL) cases, 1 Hodgkin’s lymphoma (HL) case and 1 multiple myeloma (MM) case. Among IgG4-RD patients with malignancy in this study, lung cancer (8 cases) was the most common malignancy. Patients were treated for malignancies with a variety of approaches (Table S1), including surgery, chemotherapy, radiation, traditional Chinese medicine (TCM) and supporting treatment.
As shown in Table 2, the expected total malignancies in a cohort of 602 IgG4-RD patients would be 3.347 based on general Chinese population estimates. In our study, 29 (4.82%) patients were identified as IgG4-RD accompanied by malignancy. The SPR for total malignancy compared to the general Chinese population was 8.66 (95%CI 5.84, 12.31). Among male and female IgG4-RD patients, the expected total malignancies according to general Chinese population would be 1.672 and 1.333, respectively. However, we observed 18 males and 11 females in our cohort, corresponding to SPRs of 10.77 (95%CI 6.41, 16.86) and 8.25 (95%CI 4.14, 14.66). Also, we calculated the SPRs for different malignancies. There was a significantly increased SPR for lymphoma (42.86 [95%CI 8.79, 123.88]) as listed in Table 2.
Predictive factors for malignancy in IgG4-RD patients
As shown in Table 3, odds ratios (ORs) were calculated by univariate analysis and we identified the following four variables as potential risk factors (P＜0.1): age at IgG4-RD diagnosis, eosinophil percentage, AGR and autoimmune pancreatitis. Among variables above, age at IgG4-RD diagnosis (OR 1.028 [95%CI 0.996-1.062], P=0.082), eosinophil percentage (OR 1.101 [95%CI 1.042-1.164], P=0.001) and autoimmune pancreatitis (OR 1.904 [95%CI 0.889-4.077], P=0.098) were positively correlated to malignancies in IgG4-RD patients, while AGR (OR 0.112 [95%CI 0.040-0.308], P＜0.001) was negatively correlated to malignancies. Based on univariate analysis and previous studies, we entered the following seven variables into a multivariate logistic regression model: age at IgG4-RD diagnosis, sex, serum IgG level, AGR, eosinophil percentage, serum IgG4 level  and autoimmune pancreatitis. Multivariate analysis confirmed that eosinophil percentage (OR 1.096 [95%CI 1.019-1.179], P=0.016), AGR (OR 0.185 [95%CI 0.061-0.567], P=0.002) and autoimmune pancreatitis (OR 2.400 [95%CI 1.038-5.549], P=0.041) were three independent risk factors of malignancy in IgG4-RD patients. Moreover, eosinophil percentage and autoimmune pancreatitis were positive correlation factors, whereas AGR was negatively associated with malignancy risk in IgG4-RD patients.
Development of a prediction model for malignancy risk of IgG4-RD
Based on the analyses above, four predictors were included in our final prediction model: age at IgG4-RD diagnosis, eosinophil percentage, AGR and autoimmune pancreatitis. To visualize the logistic regression model, a nomogram incorporating each of these variables was configured as shown in Figure 2. Malignancy risk assessment of a patient with IgG4-RD contains three main steps. First, determine and locate the patient’s position on each predictor axis. Second, draw perpendiculars from the corresponding axis of each predictor until the lines intersect with the top line labeled ‘Points’. Third, sum up the points for all predictors and draw a line descending from the axis labeled ‘Total points’ until it reaches the bottom line labeled ‘Malignancy risk’ to determine the probability of malignancy.
Validation of the nomogram
An internal validation was performed to test the perfomance of our nomogram using the bootstrap method with 1000 repetitions. Harrell’s C statistic was 0.738 (95%CI 0.635-0.842). Additionally, the calibration curve showed good agreement between the actual probability and the predicted probability by our nomogram (Figure 3).