Diagnostic value of serum human epididymis protein 4, carbohydrate antigen 125 and their combination in endometrial cancer: A meta-analysis.

Endometrial cancer (EC) is an epithelial malignant tumor originating from the endometrium, which is one of the 3 most common malignant reproductive system tumors in women, accounting for about 20% to 30% of the female reproductive system malignancies. [ 1 ] In recent years, the incidence of EC has been on the rise worldwide with the increase of high-risk factors such as obesity, hypertension, diabetes, and female endocrine diseases, etc. More than 90% of patients with EC have an age of onset above 50 years, and the median age of diagnosis is 63 years, while 4% of patients with EC have an age of onset below 40 years, with an increasing trend towards younger age of onset. [ 2 ] Unlike patients with ovarian cancer, most patients with EC can show symptoms such as menstrual disorders or postmenopausal uterine bleeding, making it possible for about 80% of patients to be diagnosed at an early stage, with a good prognosis compared to other gynecologic malignancies and a 5-year survival rate >95%. [ 3 ] However, 20% of patients are still asymptomatic and have local spread or distant metastases at the time of detection, losing the optimal period of treatment, with a significantly lower 5-year survival rate of 68% and 17%, respectively. [ 2 ] Therefore, improving the early diagnosis rate of EC is the key to improve the prognosis and overall survival of patients. Early screening modalities commonly used in clinical practice include pelvic ultrasound, computed tomography, magnetic resonance imaging, endometrial aspiration biopsy, diagnostic curettage, hysteroscopic diagnostic curettage, and tumor markers. For women with initial bleeding symptoms, transvaginal ultrasound is recommended and endometrial thickness (ET) will provide an important referral basis. However, CLARKE et al found that ET did not provide meaningful risk stratification: women with ET >4 mm and ≤4 mm had an almost equal risk of EC/endometrial intraepithelial neoplasia. [ 4 ] Serum tumor markers are the most widely used tumor detection tools in clinical practice. Compared with pathological biopsies or tissue biomarkers of surgical specimens, they have the advantages of convenient detection, less invasive, and easily accepted by patients. However, there are no serum tumor markers routinely used for early diagnosis, prognosis or recurrence monitoring of EC yet. Carbohydrate antigen 125 (CA125) is the most studied tumor marker in diagnosis and treatment of EC, but it may also be elevated in physiological or pathological conditions other than cancer, such as menstruation, pregnancy, endometriosis, and pelvic inflammatory diseases, etc. [ 5 ] The specificity (SPE) of screening for EC is poor and the sensitivity (SEN) is not ideal. [ 6 ] Human epididymis protein 4 (HE4) is one of the most promising tumor markers and has been approved for the diagnosis and recurrence monitoring of ovarian cancer. Notably, except for epithelial ovarian cancer, HE4 is significantly expressed in EC, while it is lowly expressed in endometriosis and other benign gynecological diseases. [ 7 ] Currently, the diagnostic value of serum HE4, CA125, and their combination is still controversial. Degez et al pooled 52 articles on HE4 and EC and found that the SEN of HE4 for the diagnosis of EC varied from 44.2% to 91% and its SPE from 65.5 to 100%, versus 24.1% to 71.5% and from 65.6% to 100% for CA125. [ 8 ] Therefore, the purpose of this meta-analysis was to evaluate the value of HE4, CA125, and their combination in the early screening of EC.

Data curation: Qi Wu, Su-ning Bai, Li-yun Song, Wen-fei Wu. Investigation: Su-ning Bai. Methodology: Qi Wu, Su-ning Bai, Li-yun Song. Software: Qi Wu, Wen-fei Wu, Lina Han. Writing – original draft: Qi Wu, Li-yun Song, Wen-fei Wu. Writing – review & editing: Qi Wu, Lina Han.

The following terms were used: “Uterine neoplasms,” “Uterine cancer,” “EC,” Endometrial neoplasms,” “Human epididymis protein 4,” “HE4,” “CA125,” “Cancer antigen 125.” “CA125,” “Cancer antigen 125” and “CA125.” The Cochrane Library, EMBASE, PubMed, Web of Science, CNKI, VIP, WanFang, and CBM were searched until November 31, 2021. EndNote X9 was applied to manage the preliminary search of relevant literature. The case group was patients with primary EC with a clear histopathological diagnosis; the control group was a healthy population or with a benign uterine disease that was easily confused with endometrium and testified to be free of any malignancy; a complete 2-by-two table on the diagnostic value of serum HE4, CA125 and their combination in EC could be extracted or calculated in the study data; blood samples were collected prior to initiation of antineoplastic therapy. Duplicate literature, reviews, cases, conference proceedings, incomplete study subgroups, or unavailable full text; combination of other malignancies; erroneous data or inability to extract complete 2-by-two table data from the literature; <30 patients with EC in the study group; literature with the quality assessment of diagnostic accuracy studies 2 (QUADAS-2) quality scores <8. All literatures were independently screened by 2 investigators, and any disagreement would be discussed or consulted with a third investigator. QUADAS-2 [ 9 ] was applied to evaluate the quality of the included literatures in 4 aspects: patient selection, index test, reference standard, and flow and timing. The QUADAS-2 quality score ≥8 was considered as high-quality literature. Two investigators independently assessed the methodological quality of the included literatures and extracted data using Review Manager 5.3. Any disagreement would be discussed or consulted with a third investigator. The extracted data consisted of: the first author name, year of publication, country, type of study design, diagnostic gold standard, mean age of the case group, and type of control group; test methods and threshold levels of serum HE4 and CA125; and 2-by-two table information extracted directly from the original literature or by calculation, including true positive, false positive, true negative, and false negative. Data extraction of the included literatures was independently performed by 2 investigators, and any disagreement would be discussed or consulted with a third investigator. Statistical analysis of the data was performed using the Midas command of Stata 16.0. The Midas command uses a bivariate mixed effects model, which is essentially a random effects model. The Q test and I 2 test were used to analyze the heterogeneity among studies. If there was no significant heterogeneity ( I 2   .1), we used a fixed effect model, otherwise a random effect model was used to calculate the pooled SEN, SPE, positive likelihood ratio (PLR), negative likelihood ratio (NLR), diagnostic odds ratio (DOR), and their 95% confidence interval (CI) for serum HE4, CA125 and HE4 + CA125, plot summary receiver operating characteristic curve (SROC) and summarize the area under the curve (AUC) to assess diagnostic accuracy. If there was significant heterogeneity, the magnitude of heterogeneity due to threshold effects was quantified, and Meta-regression analysis and subgroup analysis were performed to find the source of heterogeneity. Deek funnel plot was used to evaluate publication bias. Metaninf command was used to perform SEN analysis of the included studies. Meta-regression analysis was performed by two of HE4, CA125, and HE4 + CA125 to compare the differences in diagnostic accuracy. All statistical tests were 2-sided and P  < .05 was considered statistically significant.

A total of 25 of 1159 studies were included, [ 10 – 34 ] and the literature screening process is shown in Figure 1 . A total of 1980 cases were included in the case group, and the age of several case groups was around 50 years old, of which the minimum mean age was 32.5 years and the maximum was 66 years. The control group consisted of 2345 individuals, of which 7 studies [ 11 , 14 , 21 , 23 , 24 , 31 , 33 ] had only healthy population as controls, 7 studies [ 10 , 15 , 17 , 19 , 22 , 32 , 34 ] had only benign uterine disease as control, and 11 studies [ 12 , 13 , 16 , 18 , 21 , 25 – 30 ] had benign uterine disease combined with healthy population as control. The basic characteristics of the included studies are shown in Table 1 . Quality evaluation of the literature showed a high, unclear and low risk of bias of 13%, 18%, and 69%, respectively. High, unclear, and low applicability issues were 8%, 7%, and 85%, respectively. Selection bias was present in all included 25 studies. The literature quality evaluation is shown in Figure 2 . The basic characteristics of the enrolled studies. CA125 = carbohydrate antigen 125, CLIA = chemiluminescent immunoassay, CMIA = chemiluminescent microparticle immunoassay, ECLIA = electrochemiluminescent immunoassay, ELISA = enzyme-linked immunosorbent assay, FN = false negative, FP = false positive, HE4 = human epididymis protein 4, NR = not reported, RIA = radioimmunoassays, TN = true negative, TP = true positive. HE4 had premenopausal and postmenopausal thresholds. Flow chart of literature search. Quality assessment results of included studies based on the quality assessment of diagnostic accuracy studies 2 (QUADAS-2) tool criteria. The total heterogeneity of the 25 included studies was significant ( I 2  = 99%, P  < .001), and the exploration of the heterogeneity of each effect variable revealed high heterogeneity in pooled SEN ( I 2  = 85.38%, P  < .001), SPE ( I 2  = 93.80%, P  < .001), PLR ( I 2  = 86.50%, P  < .001), NLR ( I 2  = 82.30%, P  < .001), and DOR ( I 2  = 100%, P  < .001). However, the threshold effect was not significant (correlation = −0.27, P  = .07) and accounted for only 7% of the overall heterogeneity. Meta-regression and subgroup analysis (Table 2 ) were performed on population, study type, test method, age of case group, type of control group and cutoff value, which found that type of control group ( I 2  = 78%, P  = .01), test method ( I 2  = 93%, P  < .001), age of case group ( I 2  = 92%, P  < .001) and cutoff value ( I 2  = 96%, P  < .001) significantly affected the diagnostic accuracy. We used bivariate mixed-effects model to pool effect variables, and the pooled SEN, SPE, PLR, NLR, DOR and AUC were 0.58 (95% CI 0.52–0.63), 0.95 (95% CI 0.92–0.97), 11.57 (95% CI 6.88–19.48), 0.45 (95% CI 0.39–0.51), 25.92 (95% CI 14.84–45.26), and 0.80 (95% CI 0.76–0.83), respectively. The forest plot for SEN and SPE and the SROC curve are shown in Figures 3 and 4 A, respectively. Meta-regression and subgroup analysis of diagnostic accuracy of HE4 in EC. AUC = area under the curve, CIA = chemiluminescent assays, DOR = diagnostic odds ratio, EC = endometrial cancer, HE4 = human epididymis protein 4, NLR = negative likelihood ratio, PLR = positive likelihood ratio, SEN = sensitivity, SPE = specificity. P value of Meta-regression between healthy and those containing benign diseases. Forest plot of sensitivity and specificity of HE4. Left panel: sensitivity, right panel: specificity. Summary receiver operating characteristic curve (SROC) of HE4 (A), CA125 (B), and HE4 + CA125 (C). Sensitivity analysis of HE4 (D), CA125 (E), and HE4 + CA125 (F). Deke funnel plots of HE4 (G), CA125 (H), and HE4 + CA125 (I). CA125 = carbohydrate antigen 125, HE4 = human epididymis protein 4. There was significant heterogeneity in the 25 included studies ( I 2  = 99%, P  < .001), with significant heterogeneity in pooled SEN ( I 2  = 91.82%, P  < .001), SPE ( I 2  = 94.11%, P  < .001), PLR ( I 2  = 87.23%, P  < .001), NLR ( I 2  = 86.82%, P  < .001), and DOR ( I 2  = 100%, P  < .001). However, the threshold effect was not significant (correlation = −0.25, P  = .07), accounting for 7% of the total heterogeneity. Therefore, Meta-regression and subgroup analysis (Table 3 ) of different populations, study designs, test methods, age of case group and types of control group included in the study revealed that different test methods ( I 2  = 90%, P  < .001) and age of case group ( I 2  = 94%, P  < .001) significantly affected diagnostic accuracy. Bivariate mixed-effects models were applied to count the pooled SEN, SPE, PLR, NLR, DOR and AUC were 0.41 (95% CI 0.34–0.49), 0.91 (95% CI 0.85–0.95), 4.55 (95% CI 2.73–7.58), 0.65 (95% CI 0.57–0.74), 7.03 (95% CI 3.92–12.62), and 0.68 (95% CI 0.64–0.72). The forest plot for SEN and SPE and the SROC curve are shown in Figures 4 B and 5 , respectively. Meta-regression and subgroup analysis of diagnostic accuracy of CA125 in EC. AUC = area under the curve, CA125 = carbohydrate antigen 125, CIA = chemiluminescent assays, DOR = diagnostic odds ratio, EC = endometrial cancer, NLR = negative likelihood ratio, PLR = positive likelihood ratio, SEN = sensitivity, SPE = specificity. P value of Meta-regression between healthy and those containing benign diseases. Forest plot of sensitivity and specificity of CA125. Left panel: sensitivity, right panel: specificity. CA125 = carbohydrate antigen 125. The overall heterogeneity of the 25 included studies was significant ( I 2  = 99%, P  < .001) and there was significant heterogeneity in pooled SEN ( I 2  = 86.68%, P  < .001), SPE ( I 2  = 92.61%, P  < .001), PLR ( I 2  = 87.81%, P  < .001), NLR ( I 2  = 84.32%, P  < .001), and DOR ( I 2  = 100%, P  < .001), while the threshold effect was not significant (correlation = -0.30, P  = .09), accounting for 9% of the total heterogeneity. Meta-regression and subgroup analysis (Table 4 ) revealed statistically significant differences in the pooled effect variables for different populations ( I 2  = 77%, P  = .01) and age of case group ( I 2  = 93%, P  < .001). Using a bivariate mixed-effects model for each effect variable, the pooled SEN, SPE, PLR, NLR, DOR, and AUC were 0.67 (95% CI 0.60–0.73), 0.92 (95% CI 0.87–0.95), 8.59 (95% CI 5.32–13.86), 0.36 (95% CI 0.30–0.44), 23.80 (95% CI 13.86–40.86), and 0.85 (95% CI 0.82–0.88), respectively. The forest plot for SEN and SPE and the SROC curve are shown in Figures 4 C and 6 , respectively. Meta-regression and subgroup analysis of diagnostic accuracy of HE4 + CA125 in EC. AUC = area under the curve, CA125 = carbohydrate antigen 125, DOR = diagnostic odds ratio, EC = endometrial cancer, HE4 = human epididymis protein 4, NLR = negative likelihood ratio, PLR = positive likelihood ratio, SEN = sensitivity, SPE = specificity. P value of Meta-regression between healthy and those containing benign diseases. Forest plot of sensitivity and specificity of HE4 + CA125. Left panel: sensitivity, right panel: specificity. CA125 = carbohydrate antigen 125. We performed 2-by-two meta-regression analysis of HE4, CA125, and HE4 + CA125. The difference in diagnostic accuracy between HE4 and CA125 was statistically significant ( I 2  = 87%, P  < .001). The difference in diagnostic accuracy between HE4 and HE4 + CA125 was not statistically significant ( I 2  = 56%, P  = .10). The difference in diagnostic accuracy between CA125 and HE4 + CA125 was statistically significant ( I 2  = 91%, P  < .001). SEN analysis showed that the stability of this study was satisfactory and the results were reliable (Fig. 4 D–F). Deek funnel plots showed no publication bias ( P  = .25, P  = .79, and P  = .40) in the included studies (Fig. 4 G–I).

There are many studies on the diagnostic value of HE4 alone or in combination with CA125 in EC, and many systematic reviews and Meta-analyses have been reported, but only the Meta-analysis reported by Chen et al comprehensively compared the diagnostic value of HE4, CA125 and their combination in EC. The results of a meta-analysis that included 8 studies showed that the AUC of serum HE4, CA125, and HE4 + CA125 for the diagnosis of EC were 0.77, 0.37, and 0.83, respectively. HE4 alone or in combination with CA125 was more accurate in diagnosing EC, and HE4 + CA125 had better accuracy than HE4. [ 35 ] However, there are only 5 studies on the value of serum HE4 combined with CA125 for the diagnosis of EC, so the reliability of the conclusion needs to be further confirmed by more research data. Our study showed that the SEN of HE4 + CA125 was 0.67, which was the highest, and the SEN of CA125 was 0.41, which was the lowest, indicating that the combined detection of HE4 and CA125 could improve the detection rate of EC. However, the SEN of all 3 tests did not show outstanding efficacy, which may lead to a part of the clinical underdiagnosis rate. The SPE of all 3 tests was above 0.90, thus all 3 tests could significantly reduce the misdiagnosis rate of EC. The likelihood ratio is a more stable composite index than the SEN and SPE and not affected by the prevalence, including PLR and NLR. When PLR > 10 and NLR < 0.1, the diagnostic test was considered to present good diagnostic performance. [ 36 ] The PLR of HE4 was the highest at 11.57, and the NLR of HE4 + CA125 was the lowest at 0.36, indicating that when HE4 levels were above cutoff value, it showed a significant advantage in predicting the probability of EC, but conversely, the probability of EC still remained 45%. CA125 had no significant advantage in predicting the probability of EC, and the effect of HE4 + CA125 was moderate. DOR is an evaluation index combining SEN, SPE, and likelihood ratio, and the higher the DOR value, the higher accuracy of diagnosing EC. HE4 had the highest DOR of 25.92, slightly higher than HE4 + CA125 (DOR = 23.80), and CA125 had the lowest DOR of 7.03. Compared with CA125, HE4 alone or in combination with CA125 showed a higher diagnostic accuracy. AUC is currently considered to be a good indicator for evaluating diagnostic efficacy in diagnostic tests, with AUC values closer to 1 indicating higher diagnostic efficacy and closer to 0.5 indicating poorer diagnostic efficacy. One study had shown that the bias of AUC due to heterogeneity is <6% even in the case of significant heterogeneity. [ 37 ] The results of this meta-analysis showed that the AUC of serum HE4, CA125 and their combination for the diagnosis of EC were 0.80 (95% CI 0.76–0.83), 0.68 (95% CI 0.64–0.72), and 0.85 (95% CI 0.82–0.88), respectively, and the AUC of HE4 + CA125 was higher than that of HE4 and CA125 by 5% and 17%, respectively. Combining the results of 2-by-two Meta-regression analysis of the 3 tests revealed that HE4 alone or in combination with CA125 had a good comprehensive effect in diagnosing EC, with HE4 + CA125 having a slightly higher diagnostic accuracy than HE4, but the difference was not statistically significant. Similarly, Behrouzi et al found that the combination of HE4 with CA125 or other biomarkers showed only marginal improvements in utility. [ 38 ] The results of all 3 tests in this Meta-analysis were highly heterogeneous, but none of the heterogeneity caused by threshold effects was significant. The meta-regression and subgroup analysis showed that the sources of heterogeneity could be the following: Type of control group. By comparing AUC values in subgroup analysis of people who are healthy, with benign disease and healthy combined with benign disease, we found that all 3 tests were more accurate in identifying normal population and EC than in identifying benign and malignant disease. This result suggests that the inclusion of only healthy populations may overestimate accuracy of diagnostic tests. However, during clinical practice, the identification of benign and malignant uterine diseases is difficult, so the clinical value of HE4 in identifying benign and malignant uterine diseases needs to be further explored. Test Method. In this study, the test methods of serum HE4 and CA125 included enzyme-linked immunosorbent assay, radioimmunoassay, chemiluminescent microparticle immunoassay, chemiluminescent immunoassay, and electrochemiluminescent immunoassay, with slight differences in the technique and type of antibodies used in the different methods, and possible differences in the results. Subgroup analysis of the test methods revealed that the AUC values of serum HE4 and CA125 by chemiluminescent assays (including chemiluminescent microparticle immunoassay, chemiluminescent immunoassay, and electrochemiluminescent immunoassay) were greater than those of non-chemiluminescent assays (including enzyme-linked immunosorbent assay and radioimmunoassay), indicating that chemiluminescent assays have higher SEN. Therefore, a more sensitive chemiluminescent assays may be applied clinically to improve the accuracy of serum HE4 for the diagnosis of EC. Population. Subgroup analysis of the study population revealed that the AUC value of serum HE4 was higher in non-Asian populations (mainly European and American populations) (AUC = 0.86) than in Asian populations (mainly Chinese populations) (AUC = 0.78), suggesting that the diagnostic accuracy of HE4 is better in non-Asian populations than in Asian populations, a conclusion similar to that of Li et al. [ 39 ] Another study showed that the overall reference value of HE4 in the Chinese apparently healthy population was 105.10 pmol/L, which is slightly lower than the HE4 level (140 pmol/L) in Western apparently healthy women, [ 40 ] while the range of HE4 reference values in most of the included studies refer to the European and American populations, which may be one of the sources of heterogeneity. Age of case group. It has been shown that serum HE4 levels increase with age, and postmenopausal women have higher serum HE4 levels than premenopausal women. [ 41 ] The risk of developing EC increases with age, and more than 90% of patients develop the disease at the age of 50 years or older. Therefore, a subgroup analysis was performed to compare the AUC values of the included studies using the mean age of 50 years as the cutoff for EC patients, and it was found that the diagnostic accuracy of all 3 tests was higher in patients with a mean age of <50 years than in patients with a mean age of more than 50 years. The diagnostic value of HE4 alone or in combination with CA125 was also found to be better than that of CA125 in the group with a mean age of 50 years or older, but in the group with a mean age of <50 years, HE4 alone was comparable to CA125, with an AUC of 0.83 for both. These differences may be due to the fact that most of the included studies used a uniform reference range established by the reagent manufacturers when setting the threshold for HE4, without taking into account the effect of age on HE4. Therefore, in clinical application, it may be necessary to develop appropriate serum HE4 threshold levels for patients of different ages to improve its diagnostic accuracy for EC. Other factors. Serum HE4 levels are significantly associated with impaired renal function and are the most common cause of elevated HE4 levels in patients with benign uterine disease. [ 42 ] In addition, serum HE4 levels were 29% higher in smokers than in nonsmokers, and women with a body mass index of 30 kg/m 2 had lower serum HE4 levels than those with a body mass index of 25 kg/m 2 . [ 42 ] However, information describing the above factors was not available in most of the included studies and therefore the source of heterogeneity could not be explored further. Limitations of this study: The heterogeneity of the included studies was significant, and Meta-regression analysis was performed to find many sources of heterogeneity, so the reliability of the results needs to be further confirmed by more homogeneous studies. The majority of the included studies were retrospective and single-center studies, which may have led to selection bias. Only a few studies analyzed the diagnostic value of HE4 or CA125 in different clinical stages and pathological types, so they were not analyzed for comparison. There were no uniform norms for HE4 cutoff values in the included studies, as individualized thresholds were developed by combining stages, subtypes and physiological factors. Most of the included populations were Asian, and the conclusions lacked more data support from non-Asian populations. In recent years, in addition to biomarkers, molecular diagnosis has also been widely used in the exploration of minimally invasive early screening of EC. At present, the most studied are the PapSEEK detection using endometrial exfoliated cells for the common mutant genes and aneuploidies of EC, and the methylation detection using endometrial exfoliated cells for the common DNA methylation characteristics of EC. [ 43 , 44 ] The SEN of PapSEEK detection combined with Tao Brush uterine cavity brush to obtain endometrial cells was 93% (114/123), and the SPE was 100% (0/125). PapSEEK detection using intrauterine lavage fluid specimens combined with next-generation sequencing technology may achieve higher detection efficiency. It is conceivable that the detection of molecular diagnosis combined with CA125 and HE4 may further improve the accuracy of the diagnosis of EC. This view will be the focus of the research on the diagnosis of EC. In conclusion, serum HE4 is the most promising biomarker for the diagnosis of EC. Without considering clinical stage and pathological type, HE4 alone or combined with CA125 showed better diagnostic efficacy, and HE4 + CA125 had slightly higher diagnostic efficacy than HE4, but did not show significant advantages. In clinical application, to further improve the diagnostic accuracy of serum HE4 for EC, individualized threshold levels should be developed by taking the influence of test methods, population, and age on serum HE4 levels into account, or by combining the above factors to develop relevant algorithms. However, there was a large heterogeneity in this Meta-analysis, so the credibility of the conclusions requires further confirmation by more homogeneous, prospective, and large sample size studies. In addition, with the rise of molecular diagnosis, the combination of multiple detection methods may provide new ideas for the diagnosis of EC.