To evaluate the clinical utility of HE4 in RA patients, we recruited two cohorts of RA patients, including the discovery cohort and the validation cohort (Table 1). The discovery cohort retrospectively collected the clinical data of 70 consecutive RA inpatients who were tested for the female-tumor biomarker screening panel from Jan. 2020 to Sep. 2020 and 64 consecutive healthy controls (HCs) who went to Peking University People's Hospital (PKUPH) for annual physical examination and were tested for the female-tumor biomarker screening panel between Jan. 2020 and Oct. 2020. To verify the clinical relevance of HE4 for RA patients, we prospectively recruited 98 RA inpatients and 75 HCs and regarded those patients as the validation cohort. All RA patients were diagnosed according to 2010 American College of Rheumatology (ACR)/European League Against Rheumatism classification criteria (12). Rheumatoid Arthritis patients with cancer or history of malignant neoplasm were excluded. Healthy controls who displayed abnormality in blood differential tests, biochemistry profile tests, autoantibody profile tests or had history of systemic diseases, neoplastic, and autoimmune/autoinflammatory diseases were excluded. The study protocol was reviewed and approved by the Ethical Committee of PKUPH (Protocol number: 2019PHB244). Informed consents of discovery cohort were waived for those patients in the discovery cohort. However, all participants in the validation cohort gave written informed consent.

Demographic features and clinical and laboratory findings, including age, disease duration, swollen joint count in 28 joints (SJC28), tender joint count in 28 joints (TJC28), erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), rheumatoid factor (RF), anti-CCP antibodies, EAMs, treatments and medical history, were collected from the medical database of PKUPH. Disease Activity Score (DAS28) was assessed as previous described (13, 14).

All the tumor biomarkers in the female-tumor biomarker screening panel, including alpha fetoprotein (AFP), carcinoembryonic antigen (CEA), neuron-specific enolase (NSE), cytokeratin 19 fragment antigen 21-1 (CYFRA21-1), cancer antigen 125 (CA125), CA15-3, CA19-9, progastrin-releasing peptide (ProGRP), squamous cell carcinoma (SCC), and HE4 were determined by Roche Cobas electrochemical luminescence analyzer (Hoffmann-La Roche AG., Basel, Switzerland), according to the manufacturer's instructions. The cut-off values for positivity of these tumor markers were determined based on the manufacturer's recommendations.

The assessment of ILD was performed by two pulmonologists and two radiologists with more than 10 years of thoracic imaging experience, mainly based on the symptoms and respective abnormalities suggestive of ILD in high-resolution computed tomography (HRCT) and pulmonary function tests (PFTs). The RA-ILD was assessed semi-quantitatively based on the HRCT scans, and any indeterminate ILD were excluded from the analyses. Pulmonary function tests were determined as forced vital capacity (FVC), the median forced expiratory volume in 1 second (FEV1), and diffusing capacity of the lung for carbon monoxide (DLco). Forced vital capacity, FEV1, and DLco were presented as a percentage of the predicted values for the patient's age, sex, and height, as previously described (15). Abnormalities of PFTs were defined as predicted values of FVC <80% and DLco <70% (16).

Continuous variables were presented as mean ± standard deviation (SD) for normal distribution or median (interquartile range, IQR) for abnormal distribution. Categorical variables were shown as numbers (percentages) of the total samples. The statistical significance between groups was assessed using the Mann-Whitney U-test, Student t-test, Chi-square (χ²)-test, where it was applicable. Spearman's correlation test was used to determine the relationships between HE4 and clinical parameters. The receiver operating characteristic (ROC) curve was generated to evaluate the sensitivity, specificity and areas under the ROC curve (AUC) with the 95% confidence interval (95% CI). The optimal cutoff value for predicting the incidence of ILD was identified by calculating the Youden index. Multivariate Logistic regression analyses were used to determine the risk factors for ILD. Once a univariate statistic was generated, the multivariate model was then built using a forward selection procedure. Variables with a p-value of <0.1 in the univariate analysis were first considered as candidates for the multivariate model, then variables with a p-value of <0.05 were used in the final model, and odds ratios (ORs) were calculated with 95% CI. Data analyses were calculated using SPSS 20.0 statistical software package (SPSS Inc., Chicago, Illinois, USA) or GraphPad Prism 8 (GraphPad Software Inc.). A significant difference was defined as p <0.05.

Detailed clinical and laboratory characteristics of the two cohort are presented in Table 1. The median duration of RA patients was 8.5 years for the discovery cohort and 9.5 years for the validation cohort. The median DAS28 was 5.4 (IQR 3.9, 7.2) for RA patients in the discovery cohort and 5.2 (IQR 3.7, 6.8) for RA patients in the validation cohort. Anti-CCP antibodies were positive in 85.7% of RA patients in the discovery cohort and 88.8% of RA patients in the validation cohort. Rheumatoid factor was positive in 74.3% of RA patients in the discovery cohort and 80.6% of RA patients in the validation cohort. The prevalence of RA-ILD was 20.0% in the discovery cohort and 34.7% in the validation cohort, respectively.

The cut-off values for positivity of all tumor markers in the female-tumor biomarker screening panel are listed in Table 2. Compared to other tumor markers, HE4 displayed the highest positive rates in RA patients (50.0%), followed by CYFRA21-1 (21.4%) and CA125 (20.0%) (Table 2). No HCs were detected positive for HE4. Of interest, the prevalence of HE4 in RA patients <60 years old, RA patients aged between 60 and 70 years and RA patients over 70 years old were 22.6, 66.7, and 77.8%, respectively.

Since HE4 was the most significant marker among all the tumor markers in female-tumor biomarker screening panel, we next focus the clinical performance of HE4 in RA. We first verified these results in the validation cohort. The positive rate of HE4 was 53.1% in this cohort, which was similar to the discovery cohort (Table 2). A similar trend was also observed in RA patients in the validation cohort.

The levels of HE4 were significantly elevated in patients with RA compared to those in HCs (Figure 1A). Further, patients with RA-ILD exhibited significantly higher levels of HE4 compared to patients without RA-ILD (p < 0.0001) (Figure 1B). The levels of HE4 were significantly associated with total leukocytes number (p = 0.003), total neutrophil number (p = 0.003), ESR (p = 0.001), CRP (p = 0.003), RF (p = 0.026), and anti-CCP (p = 0.006) (Figures 2A–F).

Clinical and laboratory characteristics were compared between HE4-positive RA patients and HE4-negative RA patients. Patients in the HE4-positive group displayed similar disease duration and DAS28-ESR (Table 3). Further, HE4-positive patients showed similar laboratory characteristics, including RF positivity, anti-CCP positivity, levels of C3, levels of C4, and levels of IgG (Table 3). Of interest, the prevalence of RA-ILD in patients positive for HE4 was twice as much as that in patients negative for HE4 (28.6 vs. 11.4%, p = 0.073) (Table 3). The higher incidence of RA-ILD in HE4-positive RA was also confirmed in the validation cohort (57.7 vs. 8.7%, p < 0.001). Consistent with higher levels of HE4 in patients with RA-ILD, the levels of HE4 displayed a significantly negative correlation with DLco% (r = −0.452, p = 0.006) (Figure 3A). In addition, a negative correlation between the levels of HE4 and FVC% (r = −0.319, p = 0.062) was noticed (Figure 3B). In contrast, no significant correlations between RF or anti-CCP and DLco% or FVC% were identified (Table 4).

Receiver operating curve (ROC) analysis was utilized to characterize the clinical performance of HE4 in identifying patients with RA-ILD (Figure 4). Based on Youden index, the optimal cut-off value of HE4 for distinguishing RA-ILD from RA-non ILD was 104.3 pmol/L with an area under the curve (AUC) of 0.790 and a sensitivity and a specificity of 70.8 and 77.5%, respectively. In addition, ROC analysis was also performed to calculate the cut-off values of high-CCP and high-RF, and the optimal levels were 133.14 U/ml and 170.5 IU/ml, respectively. These cut-off values were further used to identify which RA patients are at risk of having ILD (Table 5). Univariable analysis showed that male gender (p = 0.002), old age (p = 0.012), and patients positive for HE4 (p < 0.001) were risk factor in predicting RA-ILD. In contrast, RA-non ILD patients tended to show higher SJC28 (p = 0.024) and TJC28 (p = 0.002) compare to patients with RA-ILD. Besides, patients with ILD had higher percentages of high-RF and high-anti-CCP than those without, but the differences were insignificant. To further assess the independent predictors for developing RA-ILD, a multivariate logistic regression analysis was performed. Of note, female gender (OR, 0.275; 95%CI, 0.085–0.893; p = 0.032), DAS28-ESR (OR, 0.821; 95%CI, 0.690–0.976; p = 0.026), and high levels of HE4 (OR, 9.080; 95%CI, 3.481–23.682; p < 0.001) persisted as independent risk factors for predicting RA-ILD (Table 5).