Rationality of the FIGO2023 staging for early-stage endometrial cancer, compared with the FIGO2009 staging

Study population and data collection

Inclusion criteria: consultation at the Gynecology Department, Peking University People’s Hospital (PKUPH), Beijing, China, from January 2006 to December 2016; EC diagnosed by surgical pathology; FIGO2009 stages I–II. Exclusion criteria: history of primary malignant tumours in systems other than the female reproductive system or the breast; treatment prior to surgery; loss to follow-up immediately after surgery.

Baseline information about each patient was collected, including age, height, weight, pregnancy, delivery, comorbidities, family history of cancer and CA125. All included patients underwent total hysterectomy, bilateral adnexectomy, selective pelvic and para-aortic lymph node dissection and pelvic lavage. Pathological information, including Bokhman classification, histological grade, myometrial invasion, cervical stromal invasion, LVSI, peritoneal cytology, oestrogen receptor (ER), progesterone receptor (PR), p53, Ki67 and lesion maximum diameter (LMD), was extracted from the original surgery pathology reports. Two independent gynaecological pathologists reviewed the pathology slides, and any disputes were submitted to an expert committee for final decision. A positive immunohistochemical result for ER, PR or p53 was defined as over 10% of tumour cells being moderately or strongly stained in one slide. The determination of the Ki67 index was based on the proportion of tumour cells showing positive staining.

Follow-up and definitions

All patients were followed up through the outpatient clinic or by phone after surgery. Information collected included patients’ symptoms, laboratory test results and postoperative adjuvant therapy. Recurrence and death from any cause during the follow-up were recorded. If any abnormality was detected during the follow-up, the patient was required to come to our centre for further testing, and the diagnosis of recurrence would be made if new lesions were found in pathology or were highly suggestive on imaging. Reasons for termination of follow-up included death from any cause, loss of contact and arrival at the final follow-up date (28 September 2021). Overall survival (OS) was defined as the time interval from the surgery date to death from any cause. Disease-free survival (DFS) was defined as the time interval from the surgery date to confirmation of EC recurrence.

Statistical analysis

Student’s t-test and χ² test were used for comparing continuous and categorical variables, respectively. OS and DFS were assessed using the Kaplan-Meier (K-M) survival analysis and the log-rank test. Univariate and multivariate Cox regression analyses were used to identify potential prognostic indicators, with HR and its 95% CI provided. Receiver operating characteristic (ROC) curve was drawn to show the efficiency of LMD for predicting death and recurrence. All statistical analyses were done using SPSS software (IBM Corp Released 2017. IBM SPSS Statistics for Windows, V.25.0). P<0.05 was considered statistically significant.

Data screening and stage shifts

All 687 patients were primarily surgically treated and diagnosed with EC by surgical pathology at the Gynecology Department of PKUPH from January 2006 to December 2016. We excluded patients who had a history of primary malignant tumours in systems other than the female reproductive system or the breast (n=9), who received treatment earlier than the surgery (n=21), who had been lost to follow-up immediately after the surgery (n=22) and who belonged to FIGO2009 stages III–IV (n=95). Finally, 540 patients were selected for further research (figure 1).

The 540 patients were restaged according to the FIGO2023 staging. Due to data limitations, we failed to classify ‘substantial’ versus ‘focal’ LVSI according to ‘≥5 vessels involved’. Instead, we relied on the diagnosis strings given by pathologists, such as ‘massive’, ‘extensive’ and ‘numerous’ in the description of LVSI. The cases considered ‘substantial’ LVSI in this study were all of the aggressive histological type and shifted from FIGO2009 stage IA/IB to FIGO2023 stage IIC, resulting in 0 patients categorised as FIGO2023 stage IIB. There were no cases in this study that qualified as low-grade endometrioid carcinomas confined to the uterus and ovaries, and therefore, 0 patients were categorised as FIGO2023 stage IA3. The staging was not adjusted to IAm_POLEmut or IICm_p53abn because the molecular typing of the cases was unknown.

A total of 81 stage shifts occurred (figure 2), all of which were stage elevations, including IA–IC (n=9), IA–IIC (n=50) and IB–IIC (n=22). The main shifting trends: cases of FIGO2009 IA stage diverged into FIGO2023 stages IA, IC and IIC; cases of FIGO2009 IA, IB and II stages converged into FIGO2023 IIC stage. According to the FIGO2023 staging, more patients were categorised as stage II (7%→20.37%). As shown in the baseline characteristics (table 1), body mass index, para, breast cancer history, the expression of ER/PR/p53 and Bokhman classification showed significant differences among different substages of FIGO2023, but not of FIGO2009. However, peritoneal cytology showed significant differences among different substages of FIGO2009, but not of FIGO2023.

Comparing oncological outcomes of stage I/II under two staging systems

The median follow-up time for all 540 patients was 75 months (ranging from 3 to 190 months). In total, there were 25 recurrences and 17 deaths (including 10 cancer-specific deaths). The difference of OS between new stages I and II was more significant (figure 3 and online supplemental table 1).

In univariate analyses, age ≥60 years, histological grade G3, LVSI, negative ER, negative PR, positive p53, LMD and FIGO2023 stage II were associated with OS. Among them, the HR of FIGO2023 stage II relative to FIGO2023 stage I is 4.133, while the HR is 2.344 for FIGO2009. All the factors that demonstrated statistical significance above were incorporated into the multivariate analysis. The independent risk factors influencing OS were obtained to be age ≥60 years, negative ER and LMD. As for DFS, age ≥60 years, histological grade G3, negative ER, negative PR and LMD were incorporated into the multivariate analysis, and the independent risk factors were also age ≥60 years, negative ER and LMD (table 2).

Comparing oncological outcomes of shifting subgroups at FIGO2009 stage IA and FIGO2023 stage IIC

411 patients were enrolled in FIGO2009 stage IA, of which 352 remained in FIGO2023 stage IA, 9 shifted to FIGO2023 stage IC and 50 shifted to FIGO2023 stage IIC. 85 patients were enrolled in FIGO2023 stage IIC, with 13 from FIGO2009 stage II, 50 from FIGO2009 stage IA and 22 from FIGO2009 stage IB. The K-M curves of OS in different shifting subgroups of FIGO2009 stage IA tended to separate, whereas OS and DFS of FIGO2023 stage IIC tended to overlap (figure 4).

For cases of FIGO2023 stage IIC, age ≥60 years and LMD were associated with OS in univariate analyses. LMD was the independent risk factor influencing OS, and also the only risk factor influencing DFS (table 3).

Association between prognosis and lesion size

Lesion size was closely related to prognosis in all 540 patients and in 85 FIGO2023 stage IIC patients. Using LMD as a stratification factor, we compared the OS and DFS of FIGO2023 stage I/II patients at different cut-off values (2, 3, 4, 5 and 6 cm) (online supplemental table 2). The difference between OS/DFS was statistically significant when the cut-off value was 4/5/6 cm in FIGO2023 stage II patients, suggesting that the LMD may further indicate prognosis. The ROC curves showed that the LMD=5.70 cm was the optimal cut-off value for predicting death and recurrence (online supplemental figure 1). The cut-off value of LMD=5.70 cm successfully further distinguished the OS and DFS of FIGO2023 stage IIC patients (figure 5).

Histopathology

Histopathology is a central feature of the new staging system. Tumour histological type is a significant predictor of prognosis in EC.5 In our study cohort, the main factor contributing to elevated staging in patients with early-stage EC was the aggressive histological type. Patients with early-stage EC of an aggressive histological type will be classified as stage IC if there is no myometrial invasion and stage IIC if there is. These patients were previously classified as FIGO2009 stage IA/IB. As can be seen in figure 4, patients of IA–IIC migration had significantly worse OS/DFS compared with those of IA–IA, reflecting the rationale for adding histological type into the new staging, whereas the K-M curves failed to reflect a significant trend in patients of IA–IC due to the small number (n=9). The new staging also emphasises the importance of evaluating LVSI and adopts ‘substantial’ LVSI (≥5 vessels involved) defined by the WHO 2020 report.6 Absent or focal LVSI is associated with a better prognosis, whereas substantial LVSI is associated with a worse prognosis.7 In this study, due to data limitations, the determination of whether LVSI was ‘substantial’ relied on the pathologist’s diagnosis strings, which may have resulted in staging bias. The new staging requires pathologists to record LVSI accurately. However, the interpretation of LVSI is inherently tricky. Surgical manipulation, poor fixation, tumour necrosis and mesenchymal retraction may cause misinterpretation of LVSI.8 9 More rigorous definitions and more specific interpretation criteria are urgently needed. Meanwhile, the amount of vascular involvement used to define ‘substantial’ LVSI needs further discussion.10–12 Therefore, adding LVSI in the FIGO2023 staging, which has not yet matured in pathological diagnosis, may bring limitations in clinical application.

Molecular typing

The new staging recommends complete molecular typing of all. Using biopsy specimens is sufficient; repeating in hysterectomy specimens is unnecessary.13 Since the groundbreaking introduction of molecular typing of EC by The Cancer Genome Atlas (TCGA) project in 2013,14 simplified alternatives to TCGA molecular typing have been used in clinical practice, and the prognostic value has been well documented.4 15–18 This study retrospectively included patients with no data on molecular typing, so there was no stage elevation due to p53abn and no stage reduction due to POLEmut, which is another limitation of this study. However, reporting molecular typing remains a severe challenge. Alternative immunohistochemistry markers are readily available, whereas POLE mutation analysis can be challenging for many centres. Differentiating pathogenic mutations from non-pathogenic mutations is another hurdle.19 It may be possible to skip POLE mutation analysis for histologically low-grade, low/intermediate-risk EC, but complete molecular typing can be highly advantageous for aggressive histological subtypes.15 When p53abn or MMRd are present, it is especially recommended undergoing POLE mutation analysis to avoid incorrect typing and overtreatment.20 21 Therefore, simpler and easier reporting protocols are needed to promote molecular typing globally.

Lesion size

In addition, our results suggest that LMD is also an important factor influencing the prognosis. The ROC curve analysis showed that 5.70 cm is the optimal cut-off value for predicting death and recurrence. Patients with different prognoses can be differentiated within FIGO2023 stage IIC when the cut-off value is 5.70 cm. Although neither of the FIGO 2009/2023 staging systems incorporates lesion size, several studies support the association of lesion size with prognosis. The Mayo criteria classify patients with endometrioid carcinoma, histological grades 1–2, myometrial infiltration <50%, tumour diameter ≤2 cm and no evidence of tumour outside of the uterine corpus as a low-risk group for those who do not require lymphadenectomy.22 Sozzi et al analysed data from 1166 patients and found that tumour size was an independent prognostic factor for local recurrence in low-risk EC and used the LMD ≥2.5 cm as the cut-off value for predicting local recurrence in patients with low-risk EC.23 Our centre conducted a study using the ultrasound results before hysteroscopy or curettage to determine the initial lesion size, then analysed its correlation with lymph node metastasis and recurrence in patients with EC. The study found that the initial lesion size ≥4.25 cm was significantly associated with the prognosis of EC.24 The staging system could also include the lesion size as a prognostic factor. However, the method and threshold for determining the lesion size still need to be discussed.

This study has several limitations. Due to small number of cases and lack of data such as lymph node ultrastaging, we did not discuss stage III/IV. Lack of molecular typing data prevented staging adjustment. LVSI evaluation criteria inconsistent with WHO 2021 may have caused staging bias. Results may be flawed due to retrospective nature. Data from a single centre may limit generalisability to other countries/regions.

In conclusion, compared with the traditional FIGO2009 staging based on tumour extent, the FIGO2023 staging incorporates complementary histopathological and molecular features to improve the prognostic value and help achieve precision medicine.25 26 However, the excessive changes and complicated factors of the new staging system pose challenges to clinicians, pathologists and epidemiologists, which may affect its global dissemination. A prudent discussion with current evidence and local medical resources is needed before adopting the new staging as an alternative to the old one.