Quantitative dynamic contrast-enhanced parameters and intravoxel incoherent motion facilitate the prediction of TP53 status and risk stratification of early-stage endometrial carcinoma

This prospective study was complied with ethical committee standards and approved by the ethics committee of the First Affiliated Hospital of Xinxiang Medical University (NO. EC-022-002) and informed consent was taken from all individual participants. From January 2021 to April 2022, 114 female patients underwent pelvic MRI due to suspected EC by clinical examination, ultrasound (US), or computed tomography (CT). Forty participants were excluded during this study: 1) 7 patients were diagnosed with an endometrial polyp, atypical hyperplasia, or other non-EC diseases; 2) 16 patients had FIGO stage ≥ II; 3) 4 patients received radiotherapy or neoadjuvant chemotherapy; 4) 3 patients had claustrophobia or other diseases that prevented them from completing all the sequences; 5) 6 patients had inadequate DCE-MRI or IVIM imaging quality for analysis due to severe artifacts, and 6) 4 patients decided to perform histological analysis and treatment in other institutes. Ultimately, 74 patients were enrolled in the study (Figure 1).

FIGURE 1.

A 1.5 T MR system (Optima MR360, Waukesha, WI, USA) with a 12-channel phased-array body coil was used in this study. The imaging protocol included oblique axial (perpendicular to the long axis of the uterus) T1WI, T2WI, DWI, IVIM, and DCE-MRI. For DWI and DCE-MRI sequences, the scans covered the anterior superior iliac spine to the symphysis pubis. For IVIM (b = 0, 20, 40, 80, 160, 200, 400, 600, 800, and 1000 s/mm²), to minimize scan time, the scan was limited to the lesion area (determined by an experienced radiologist from the DWI images), and its location, layer thickness, and layer spacing were consistent with the corresponding layer of DWI.¹⁸ DCE-MRI was performed by a three-dimensional liver acquisition with volume acceleration (3D-LAVA) sequence with 40 phases (time resolution, 9s), and gadopentetate dimeglumine (Gd-DTPA, Bayer Pharmaceutical, Berlin, Germany) was injected intravenously with an automatic injector (0.2 mL/kg, 3.0 mL/s). The protocol details are provided in Table 1.

All images were transferred to the Advantage Workstation (version 4.7), and the IVIM and DCE-MRI images were analyzed within the workstation using vendor-provided software named MADC and GenIQ, respectively. The IVIM parameters were calculated by the following formula: [1] Sb/S0=1−f×exp−b×D+f×exp−b×D* {{\rm{S}}_{\rm{b}}}/{{\rm{S}}_{\rm{0}}} = \left( {1 - {\rm{f}}} \right) \times {\rm{exp}}\left( { - {\rm{b}} \times {\rm{D}}} \right) + {\rm{f}} \times {\rm{exp}}\left( { - {\rm{b}} \times {\rm{D}}*} \right) where S₀ was the signal intensity at the b value of 0; S_b was the signal intensity at the b value denoted by the subscript; D was the true diffusion coefficient of a water molecule; D* was the pseudo-diffusion coefficient due to microcirculation; and f was the microvascular volume fraction, indicating the fraction of diffusion related to microcirculation.¹¹ The DCE-MRI perfusion parameters were quantitatively calculated based on the Tofts model. The arterial input function (AIF) was obtained from the internal iliac artery. The imaging parameter K^trans, known as the volume transfer constant, represents the diffusion of contrast medium from the vessel to the extravascular extracellular space (EES); K_ep, known as the rate transfer constant, represents the diffusion of contrast medium from the EES to the vessel; and V_e represents the volume of EES per unit volume of tissue ¹³; thus, K_ep = K^trans/V_e.

For regions of interest (ROI), first, images of DCE-MRI and IVIM were co-registered, and then on the DCE-MRI images of the phase with the clearest lesion display²¹, ROIs were delineated layer by layer for all slices containing the tumor, and these ROIs were manually drawn along the inside margin of the primary tumor, avoiding areas with cystic degeneration, necrosis, apparent signs and hemorrhage artifacts, and blood vessels. Subsequently, all completed ROIs were automatically copied to the pseudo-color maps of the DCE-MRI and IVIM-derived parameters to calculate the mean values based on the volume of interest (VOI). All of these procedures were completed independently by two radiologists with 7 and 15 years of experience who were blinded to each other's results and the patient's clinicopathological data.

All lesion specimens were obtained surgically, and the median interval from pelvic MRI examination to surgery was 12 days (1–25 days). The specimens were processed by an experienced pathologist. The histological subtype, grade, and LVSI were confirmed by hematoxylin/eosin (HE) staining. The stage was estimated with the FIGO staging system.²² According to the European Society for Medical Oncology (ESMO) clinical practice guidelines, low-risk patients were classified into the low-risk group, while intermediate-risk, high-intermediate-risk, and high-risk patients were classified into the non-low-risk group.⁴ The TP53 status was evaluated by immunohistochemical (IHC) staining, where non-staining was viewed as the wild group, and faint, moderate, and strong staining was viewed as the mutant group. Ultimately, risk stratification was evaluated in all 74 patients, and TP53 status was evaluated in 46 patients (28 patients declined IHC for financial or other reasons).

All data were analyzed with Stata version 16.0 (Stata Corp) and MedCalc version 15.0 (MedCalc Software). P < 0.05 was considered statistically significant. The interobserver consistency of two radiologists was classified using the intraclass correlation coefficient (ICC) as poor (ICC < 0.40), fair (0.40 ≤ ICC < 0.60), good (0.60 ≤ r < 0.75), or excellent (ICC ≥ 0.75).²³ The Shapiro–Wilk test was employed to check the normality of the data. The Mann–Whitney U test and the independent samples t-test were used for nonnormally distributed data (median and interquartile range) and normally distributed data (mean ± standard deviation), respectively. The area under the receiver operating characteristic (ROC) curve (AUC) was employed to quantify the diagnostic efficacy of different parameters, and the differences were assessed using DeLong analysis. The combination of parameters was investigated by logistic regression, evaluated by bootstrap (random number set 123, repeated sampling 1000 times, backward strategy, bounded by a value of 0.1), calibration curves, and decision curve analysis (DCA).²⁴

The clinicopathological and imaging characteristics are shown in Table 2 and Figure 2, respectively.

FIGURE 2.

The D, D*, f, K^trans, V_e, and K_ep measured by 2 radiologists had excellent consistency, and the ICCs were 0.864 (95% CI: 0.788 – 0.913), 0.799 (95% CI: 0.696 – 0.867), 0.855 (95% CI: 0.729 – 0.918), 0.868 (95% CI: 0.799 – 0.915), 0.834 (95% CI: 0.748 – 0.892), and 0.828 (95% CI: 0.739 – 0.888), respectively. The average results were used for the ultimate analysis.

The K^trans and K_ep were higher and D was lower in the TP53-mutant group than in the TP53-wild group (P = 0.038, 0.002, and 0.037, respectively), f, D*, and V_e were not significantly different between the two groups (P = 0.750, 0.604, and 0.434, respectively). The K^trans, V_e, f, and D values were lower in the non-low-risk group than in the low-risk group (P < 0.001, < 0.001, 0.002, and < 0.001, respectively), K_ep and D* were not significantly different between the two groups (P = 0.218 and 0.601) (Table 3, Figure 3).

FIGURE 3.

In the identification of TP53-mutant and TP53-wild early-stage EC, the potential related factors such as age, tumor size, risk stratification, FIGO stage, subtype, grade, LVSI, D, D*, f, K^trans, V_e, and K_ep were all enrolled in regression analysis. Univariate analysis demonstrated that grade, D, K^trans, and K_ep were all risk predictors (P all < 0.1), while multivariate analysis showed that only D and K^trans were independent predictors (P = 0.003, 0.016).

In the identification of non-low-risk and low-risk early-stage EC, potential risk-related factors such as age, tumor size, TP53 status, D, D*, f, K^trans, V_e, and K_ep were all enrolled in regression analysis. Univariate analysis demonstrated that tumor size, D, f, K^trans, and V_e were all risk predictors (P all < 0.1), while multivariate analysis showed that only f, K^trans, and V_e were independent predictors (P = 0.036, 0.003, and 0.024, respectively) (Table 4).

In the differentiation of TP53-mutant and TP53-wild early-stage EC, the combination of independent predictors (K^trans and D) showed the optimal diagnostic efficacy (AUC = 0.867; sensitivity, 92.00%; specificity, 80.95%; P < 0.001), which was significantly better than D (AUC = 0.694, Z = 2.169, P = 0.030), and K^trans (AUC = 0.679, Z = 2.572, P = 0.010). However, the difference between the combination of independent predictors and K_ep (AUC = 0.773) was not significant (AUC = 0.773, Z = 1.272, P = 0.203) (Figure 4A, Table 5).

FIGURE 4.

In the differentiation of low-risk and non-low-risk early-stage EC, the combination of independent predictors (f, K^trans, and V_e) showed the optimal diagnostic efficacy (AUC = 0.947; sensitivity, 83.33%; specificity, 93.18%; P < 0.001), which was significantly better than D (AUC = 0.761, Z = 3.113, P = 0.002), f (AUC = 0.688, Z = 4.317, P < 0.001), K^trans (AUC = 0.852, Z = 2.713, P = 0.007), and V_e (AUC = 0.808, Z = 3.175, P = 0.002) (Figure 4B, Table 5).

Bootstrapped samples were used to validate the combination of independent predictors. The ROC and the calibration curve indicated that the validation models not only had high accuracy in identifying TP53-mutant and TP53-wild early-stage EC (AUC, 0.815; 95% CI, 0.782 – 0.846, Figure 5A), and low-risk and risk early-stage EC (AUC, 0.922; 95% CI, 0.895 – 0.940, Figure 6A), but also highly had good consistency (Figure 5B, 6B). Also, DCA showed that the above combinations of independent predictors were both reliable clinical decision tools (Figure 5C, Figure 6C).

FIGURE 5.

FIGURE 6.

The parameter D of IVIM can reflect the diffusion movement of water molecules in the tissue, and usually, the more obvious the restriction of water molecule diffusion, the smaller the D value.¹³ In this study, the D value of the TP53-mutant group was significantly lower than that of the TP53-wild group, which was similar to the results of Wang et al. in the field of epithelial ovarian cancer²⁵, suggesting that D values can be used to predict TP53 status of early-stage EC. Presumably, the reason was that TP53-mutant has a faster rate of cell proliferation than TP53-wild, which easily impedes the diffusion of water molecules, resulting in a lower D value.²⁶ In addition, D could also be used to assess the risk stratification of early-stage EC in the present study, which was consistent with previous studies.^17,19 The reason may be that there were differences in histological grade, FIGO stage, and lymph node metastasis between the low-risk and non-low-risk early-stage EC, resulting in different degrees of influence on the diffusion of water molecules and ultimately leading to significant differences in D values between the two groups.^18,20

D* was a perfusion parameter of IVIM that is mainly correlated with the velocity of blood flow within the microcirculation.¹³ Previous publications have demonstrated that D* values with poor stability and repeatability could not effectively evaluate histopathological information of early-stage EC due to the influence of the scanning parameters, the ROI determination method, the signal-to-noise ratio (SNR), and other factors.^17,18,19,20 In this study, there was no statistically significant difference in D* between the TP53-mutant and TP53-wild groups, and the low-risk and the non-low-risk groups, which was consistent with the above research, further proving that the D* value was unable to play a role in the assessment of TP53 status and risk stratification in early-stage EC.

As another perfusion parameter derived from IVIM, f was mainly related to the microvascular density of the tissue.^13,27 A study by Zhang et al. involving 53 participants showed that although high-risk early-stage EC is metabolically active and rich in neovascularization, due to the dense tissue structure and more necrotic tissue, its overall internal microvascular density is instead reduced compared to that of low-risk early-stage EC, so the f value decreases.²⁰ This trial was conducted on a larger sample size of patients (n = 74) and obtained results consistent with those of Zhang et al.¹⁹ Further analysis also identified the f value as an independent predictor for discriminating between low-risk and non-low-risk early-stage EC. However, there were also studies that have shown conflicting results of f values in the assessment of lesions. For example, the study by Meng et al. showed that high-risk early-stage EC had higher f values than low-risk early-stage EC.²⁰ Similarly, in the assessment of gliomas, the study of Bai et al. showed that low-grade gliomas had higher f values than high-grade gliomas¹⁴, while Shen et al. concluded that high-grade gliomas have greater f values.¹⁶ We speculate that the above phenomenon may be caused by the variations in scanning equipment and b-value settings²⁸, as well as the shortcoming that the f value itself is susceptible to T2 contribution and relaxation effects.²⁹ In addition, the results of this study also showed that the f was similar to D* and could not differentiate TP53-mutant from TP53-wild early-stage EC, which to some extent suggests that the use of IVIM perfusion parameters to assess the TP53 status of early-stage EC may still need further exploration.

K^trans is the most significant perfusion-related parameter in DCE-MRI, mainly reflecting the transfer rate of the contrast agent from the vessel to the EES.³⁰ Previous studies have shown that the more neovascularization in the tissue and the greater the permeability, the greater the K^trans value.31 In terms of TP53 status assessment, the present study found a significantly higher K^trans value in the TP53-mutant group compared with the TP53-wild group, which we suggest may be related to the ability of TP53 gene overexpression to promote angiogenesis.^2,3 In terms of risk stratification assessment, several studies have shown that EC with aggressive characteristics, such as grade 3, advanced FIGO stage, and non-endometrioid sub-type, grows quickly without sufficient neoangiogenesis (i.e., blood support), resulting in tissue hypoxia. Hypoxia will lead to tissue necrosis and the formation of hypoperfused areas, thus eventually causing a decrease in overall tumor perfusion and a decrease in K^trans values.12,17,32,33 In this work, the K^trans value was significantly lower in the non-low-risk group than in the low-risk group, which was consistent with the above findings and further demonstrates that the K^trans value can play a role in the risk stratification of early-stage EC.

K_ep was designed to reflect the transfer rate of the contrast agent from the EES into vessels, so similar to K^trans, its size was closely related to the number of new vessels and vascular permeability.²⁹ In this study, since TP53 overexpression can promote angiogenesis^2,3, the K_ep value of the TP53-mutant group was significantly higher than those of the TP53-wild group, and the diagnostic efficacy was 0.773. However, the K_ep value did not show significant value in the identification of different risk stratifications, which was not consistent with the study of Ye et al.¹² We speculated that this may be because the study by Ye et al. included both early-stage (stage I) and advanced-stage (stage II, III, and IV) EC, whereas the present study population included only early-stage EC, which reduced the differences in patients between the different groups and ultimately resulted in nonfunctional K_ep values.

V_e is a parameter in DCE-MRI that can reflect the volume of EES. In the present study, there was no significant difference in V_e between the TP53-mutant and TP53-wild groups, which may be related to the fact that TP53 overexpression promotes both cell proliferation and angiogenesis, resulting in difficulty in significant changes in EES.^2,3 In terms of risk stratification assessment, V_e values in the non-low-risk group were significantly smaller than those in the low-risk group, which was similar to the results of previous studies^17,34, and we speculated that the reason for this result may lie in the fact that the non-low-risk group had greater invasiveness and therefore greater cell density, tighter tissue structure, and smaller EEC compared with the low-risk group. However, some studies have also concluded that V_e was difficult to use in the evaluation of diseases such as EC and breast cancer.^12,35 This may be related to the fact that Ve is less stable and susceptible to factors such as lesion edema and microcystic changes.³⁶ In a follow-up study, we will expand the sample size and further explore the role of V_e in EC assessment to obtain more convincing results.

The diagnostic efficacy of the combination of independent predictors and each individual parameter was compared in this study, and the results showed that the diagnostic efficacy of the former was significantly higher than that of the latter, which may be because the combination of independent predictors concentrates the advantages of different parameters and therefore can reflect the lesion characteristics more comprehensively and accurately. Therefore, we suggest that the combined application of IVIM and DCE-MRI in clinical routine may provide a more reliable basis for the TP53 status and risk stratification prediction of early-stage EC when conditions permit.

In this study, TP53 mutation and risk stratification in early-stage EC were included in each other's regression analysis, and the results showed that neither was a predictor of the other. Although the small sample size may have affected the reliability of the above results to a certain extent, it indicates to some extent that the TP53 status in early-stage EC is not significantly correlated with risk stratification. In the future, as the sample size increases, we will conduct more in-depth studies on the relationship between the two, with a view to obtaining more accurate results.

This study has several limitations. First, our study was designed at a single institution with a relatively small number of patients, especially since some patients forgo immunohistochemical testing for financial reasons, which may have led to selection bias. Second, due to the small sample size, this study did not set up a separate validation set but used the bootstrap (1000 samples) method to validate the combination of independent predictors, which may have reduced the reliability of the experimental results. Third, areas of cystic degeneration, necrosis, apparent signs and hemorrhage artifacts, or vessels were avoided in the delineation of the ROI, which may influence the determination of some parameters. Finally, the machine used in this study was a 1.5 T MRI, and its imaging quality and parameter reliability may be inferior to those of a 3.0 T MRI.