ABSTRACT
Objective
This study aimed to construct a nomogram incorporating preoperative laboratory parameters and clinical pathological factors for the first time to predict the probability of adjuvant therapy requirement following radical surgery in IB stage cervical squamous cell carcinoma (SCC) with tumor size ≤ 4 cm.
Methods
Clinical pathological parameters and relevant laboratory indicators were collected from IB stage cervical SCC with tumor size ≤ 4 cm patients who underwent radical surgery at our hospital. Patients included in the study were randomly divided into a training set and a validation set in a 7:3 ratio. In the training set, the least absolute shrinkage and selection operator (LASSO) regression and multivariate logistic regression were used to determine the final variables for constructing the nomogram. Finally, the performance of the nomogram was evaluated using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA) in both the training and validation sets.
Results
The nomogram ultimately included seven predictive variables. The area under the ROC curve (AUC) for the nomogram in the training and validation sets was 0.863 and 0.767, respectively. Moreover, the calibration curve of the nomogram was relatively close to the ideal curve. The DCA showed that using the nomogram to predict the probability of adjuvant therapy requirement over a wide range of threshold values is beneficial.
Conclusion
This study constructed and validated a model for predicting adjuvant therapy requirement in IB stage cervical SCC with tumor size ≤ 4 cm. The model can help clinicians determine the risk of postoperative adjuvant therapy before surgery, promoting personalized treatment choices and patient management.
Keywords: adjuvant therapy, cervical squamous cell carcinoma, nomogram, predictive model
1. Introduction
Cervical cancer is the fourth most common cancer among women worldwide and the fourth leading cause of cancer‐related deaths in women [1]. In 2020, there were approximately 600,000 new cases of cervical cancer and 340,000 deaths due to cervical cancer globally [2]. Squamous cell carcinoma (SCC), adenocarcinoma (AC), and adenosquamous carcinoma (ASC) are the three most common histological types of cervical cancer, with SCC accounting for about 80% of all cervical cancers [3].
For patients with IB stage cervical cancer with tumor size ≤ 4 cm, the initial treatment options include surgery or radiotherapy, with no difference in outcomes. If radical hysterectomy is chosen, supplementary treatment may be required based on high‐risk and/or intermediate‐risk factors identified in the postoperative pathological findings [4]. Postoperative pathological results for IB stage cervical cancer with tumor size ≤ 4 cm patients indicate a lymph node metastasis rate of about 19% [5], and when combined with other high‐risk and/or intermediate‐risk pathological factors requiring supplementary treatment, a significant proportion of IB stage with tumor size ≤ 4 cm patients undergo radical surgery followed by postoperative adjuvant therapy (chemoradiation). However, the therapeutic benefit of such adjuvant therapy has been increasingly questioned. The NRG Oncology/GOG‐263 randomized Phase III trial demonstrated that adding concurrent chemotherapy to adjuvant radiation in intermediate‐risk early‐stage cervical cancer did not significantly improve recurrence‐free survival compared with radiation alone (3‐year RFS: 88.5% vs. 85.4%, p = 0.09), while substantially increasing grade 3–4 adverse events (43% vs. 15%) [6]. Similarly, a propensity score‐matched analysis by Tanaka et al. suggested that adjuvant chemotherapy alone may be superior to concurrent chemoradiotherapy in specific subgroups of T1b cervical cancer [7], and a recent systematic review and meta‐analysis indicated that routine adjuvant therapy after radical hysterectomy may not significantly reduce recurrence or mortality in all early‐stage cervical cancer patients [8]. Radical surgery followed by postoperative adjuvant therapy leads to increased complications and increases the financial burden of patients. Specifically, there are the following two aspects: First, this treatment sequence exposes patients to the dual risks of surgery and supplementary treatment (radiotherapy, chemotherapy, chemoradiotherapy), and surgery carries the risks of damaging blood vessels, nerves, the urinary system (ureter, bladder, etc.), and the digestive system (rectum, etc.) [9], and a considerable proportion of patients develop lymphocele [10], bladder dysfunction [11], and other complications after surgery. Postoperative supplementary treatment can lead to radiation enteritis [12], radiation cystitis [13], radiation peripheral neuropathy [14], and radiation hematopoietic system damage [15], and studies have shown that the incidence of lower limb lymphedema, renal insufficiency, and diarrhea in patients requiring postoperative adjuvant therapy following radical surgery is significantly higher than in those undergoing radical hysterectomy alone [16]. Second, these patients need to bear the dual costs of surgery and supplementary treatment, increasing the financial burden and poorer health economic benefits. Therefore, there is an urgent need for new diagnostic methods or predictive models to reduce the need for postoperative adjuvant therapy following radical surgery in these patients.
Currently, clinicians primarily rely on FIGO 2018 staging and preoperative imaging to predict the risk of postoperative adjuvant therapy. However, these conventional clinical indicators alone demonstrate limited predictive accuracy for postoperative adjuvant therapy requirement, as they cannot fully capture the tumor's biological aggressiveness and host systemic status. Emerging evidence suggests that laboratory‐derived inflammatory and nutritional indices reflect the tumor microenvironment and host immune response, which are closely associated with tumor progression and treatment outcomes in cervical cancer [17, 18, 19, 20, 21, 22, 23, 24, 25, 26]. Based on this, some scholars have combined laboratory indicators with clinical pathological parameters to construct a predictive model for pelvic lymph node metastasis in early‐stage cervical cancer [27, 28], all of which have shown good discrimination and accuracy. However, emerging approaches have incorporated advanced imaging techniques; for instance, Meng et al. developed a CT‐based preoperative nomogram for lymph node metastasis prediction [29], while Yang et al. proposed a novel ultrasound radiomics model with improved diagnostic performance [30]. Despite these advances, all of which have shown good discrimination and accuracy, there remains no predictive model specifically designed for postoperative adjuvant therapy in cervical cancer. Therefore, integrating readily available laboratory parameters with clinical pathological factors may compensate for the deficiencies of existing staging systems and improve preoperative prediction of adjuvant therapy requirements.
In this study, we conducted a retrospective analysis of the IB stage cervical SCC with tumor size ≤ 4 cm cohort at our hospital, and for the first time constructed and validated a model for predicting adjuvant therapy requirement in IB stage cervical SCC with tumor size ≤ 4 cm. We proposed that this prediction model be readily accessible to clinicians, helping to reduce the need for postoperative adjuvant therapy, decrease complications, and improve health economic benefits.
2. Patients and Methods
2.1. Patients
This study retrospectively included patients with IB stage cervical SCC with tumor size ≤ 4 cm who underwent radical surgery at our hospital from October 2016 to October 2018. The study was reviewed and approved by the hospital's ethics committee, which waived the requirement for informed consent in accordance with the Declaration of Helsinki, and all personal information of the patients involved in this study was anonymized. All surgeries were performed by experienced gynecologic oncologists at our hospital, using either open or laparoscopic approaches, with the surgical scope including at least radical hysterectomy and pelvic lymph node dissection. The study flowchart is shown in Figure 1.
FIGURE 1.
Flowchart of this study.
2.2. Definition of Postoperative Adjuvant Therapy
Postoperative adjuvant therapy is defined as the need for supplementary treatment following radical surgery. Supplementary treatment can be radiotherapy alone (intracavitary radiotherapy, external irradiation) or chemotherapy alone, or a combination of radiotherapy and chemotherapy (concurrent chemoradiotherapy). The specific plan is determined based on the high‐risk and intermediate‐risk pathological factors of the patient after the operation. The decision of postoperative supplementary treatment for patients is based on the pathological findings, with at least one high‐risk pathological factor (positive pelvic lymph nodes, positive surgical margins, parametrial invasion) or intermediate‐risk pathological factor (lymphovascular space invasion, deep stromal invasion, large primary tumor) meeting the Sedlis criteria [31].
2.3. Data Collection
In this study, we used our Hospital Information System (HIS) to collect patient data obtained through relevant literature and clinical judgment, including general conditions, preoperative pathological data, and laboratory data. General conditions of the patients included: age at diagnosis, age at menarche, age at first sexual intercourse, main symptoms, body mass index (BMI), gynecological examination, and imaging examinations. Pathological data included only the degree of histological differentiation. Laboratory data included preoperative blood routine, coagulation function, biochemical tests, and tumor markers. Immediately after data extraction, we carried out rigorous data cleaning and consistency checks, which included the following three steps. First, we performed range and logic checks on all variables to identify and exclude any obvious outliers that fell outside of the expected range or defied logical reasoning. Next, we excluded patients with missing data to ensure that all patients included in the study had a reliable source of data. Finally, in order to further ensure data accuracy, we conducted a sampling reconciliation of the data in the HIS with patients' paper records.
2.4. Potential Predictive Factors and Data Processing
The selection of potential predictors was based on three criteria: (i) established prognostic relevance in cervical cancer or other gynecological malignancies; (ii) biological plausibility linking the marker to tumor progression, inflammation, or host nutritional status; and (iii) routine availability in preoperative laboratory panels to ensure clinical practicability. Specifically, total cholesterol (TC) and triglycerides (TG) were included because recent studies have identified dyslipidemia as a metabolic hallmark of cancer progression. For instance, Lin et al. (2021) reported that serum cholesterol and triglycerides were predictive of survival in cervical cancer patients [19]. Lipid metabolism reprogramming supports rapid tumor cell proliferation and membrane synthesis, and low cholesterol levels may reflect advanced disease status or malnutrition. Although their direct association with adjuvant therapy requirement has not been specifically investigated, their prognostic value in cervical cancer justified their inclusion as exploratory candidates in this predictive model.
The impact of tumor size and squamous cell carcinoma antigen (SCC‐Ag) on the prognosis of cervical SCC is well established, and the histological grading (degree of differentiation) of cervical cancer is also correlated with prognosis [32]. In terms of clinical manifestations, patients with abnormal uterine bleeding may be associated with larger volume, more aggressive tumors, and thus related to prognosis. Many recent studies have shown that laboratory indicators and their derived indices are related to the prognosis of cervical cancer [17, 18, 19, 20, 21, 22, 23, 24, 25, 26]. In summary, this study defined tumor size, clinical manifestations, degree of differentiation, SCC‐Ag, white blood cell count (W), neutrophil percentage (N%), lymphocyte percentage (L%), neutrophil count (N), monocyte count (M), lymphocyte count (L), neutrophil‐to‐lymphocyte ratio (NLR), monocyte‐to‐lymphocyte ratio (MLR), platelet‐to‐lymphocyte ratio (PLR), systemic immune‐inflammation index (SII), red blood cell count (R), hemoglobin (Hb), platelet count (Plt), prothrombin time (PT), international normalized ratio (INR), activated partial prothrombin time (APTT), fibrinogen (Fg), thrombin time (TT), alanine aminotransferase (ALT), aspartate aminotransferase (AST), albumin (ALB), globulin (GLB), Onodera prognostic nutritional index (OPNI), lactate dehydrogenase (LDH), TC, TG, alkaline phosphatase (ALP), albumin/alkaline phosphatase ratio (ALB/ALP), albumin/fibrinogen ratio (ALB/Fg) and Hemoglobin, Albumin, Lymphocyte, and Platelet (HALP) score as potential predictive factors, where SII = PLT × N ÷ L, OPNI = ALB (g/L) + 5 × L (10 [9]/L), HALP score = Hb × ALB × L ÷ Plt. Receiver operating characteristic (ROC) curve was used to select continuous variables related to postoperative adjuvant therapy, and logistic regression was used to assess the correlation between categorical variables and postoperative adjuvant therapy. The ROC curve and Youden index were subsequently used in the training cohort to determine the optimal cut‐off values for continuous variables, which were then transformed into binary variables and applied to both cohorts.
2.5. Statistical Analysis
Patients included in the study were randomly divided into a training set and a validation set in a 7:3 ratio and compared using Pearson's chi‐square test. The 7:3 ratio for splitting training and validation sets was chosen based on established conventions in predictive modeling research, balancing the need for adequate sample size in the training cohort to ensure model stability while retaining sufficient cases in the validation cohort for reliable performance assessment. Regarding sample size adequacy, with 273 patients in the training set (including 113 postoperative adjuvant therapy cases) and 117 in the validation set (including 46 postoperative adjuvant therapy cases), the event‐per‐variable (EPV) ratio exceeded 10:1 (113 events / 8 candidate variables = 14.1), which satisfies the widely accepted minimum EPV > 10 criterion for reliable logistic regression model development. Additionally, the total sample size of 390 patients provides adequate power to detect meaningful differences, as confirmed by our stable 10‐fold cross‐validation results (mean accuracy 0.80, SD 0.04).
In the training set, the least absolute shrinkage and selection operator (LASSO) regression analysis was used to select predictive variables for the nomogram. The selected variables were included in subsequent univariate and multivariate logistic regression, and the nomogram for adjuvant therapy requirement was constructed based on the results of the multivariate logistic regression. The performance of the nomogram was assessed using ROC curves and calibration curves, with both discrimination and calibration assessed by bootstrapping with 1000 resamples. Decision curve analysis (DCA) was performed to determine the net benefit threshold values for the nomogram. We used 10‐fold cross‐validation to further verify the model's performance. The dataset was randomly divided into 10 subsets. Each time, 9 subsets were used to train the model, and the remaining 1 subset was used for validation. This was repeated 10 times to ensure each subset served as the validation set. We calculated the average of the 10 validation results to assess the model's stability and generalization ability across different data subsets. Results with a p‐value less than 0.05 were considered significant, and all statistical analyses were performed using R software (version 4.2.2).
3. Results
3.1. Patient Characteristics
The baseline characteristics of the patients are shown in Table S1. A total of 390 patients were included in this study, of whom 159 required postoperative adjuvant therapy.
3.2. Preliminary Screening of Predictive Factors
ROC curves (Figure S1) were plotted for the continuous variables included in the study, and the area under the curve (AUC) (Table S2) was calculated. It was found that tumor size (AUC 95% CI: 0.733–0.823), SCC‐Ag (AUC 95% CI: 0.636–0.744), APTT (AUC 95% CI: 0.519–0.634), Fg (AUC 95% CI: 0.511–0.627), ALT (AUC 95% CI: 0.527–0.641), and ALB/Fg (AUC 95% CI: 0.523–0.638) had AUC > 0.5, and these variables were included in subsequent studies; while BMI (AUC 95% CI: 0.495–0.611), SII (AUC 95% CI: 0.498–0.614), OPNI (AUC 95% CI: 0.495–0.613), TC (AUC 95% CI: 0.495–0.611), and HALP score (AUC 95% CI: 0.497–0.615) had the minimum AUC ≥ 0.495, close to the critical value (0.5), and were therefore also included in subsequent studies.
Logistic regression between the categorical variables included in the study (clinical manifestations and degree of differentiation) and postoperative adjuvant therapy requirement was shown in Table S3, with p value all less than 0.05, so they were included in the subsequent study.
3.3. Splitting the Training and Validation Sets
As shown in Table 1, the patients included in the study were randomly divided into a training set (n = 273) and a validation set (n = 117) in a 7:3 ratio. 113 patients in the training set required postoperative adjuvant therapy, and 46 patients in the validation set required postoperative adjuvant therapy. Statistical analysis showed that there were no significant differences in the predictive factors screened in the preliminary stage between the training set and the validation set (p > 0.05).
TABLE 1.
Variables after preliminary screening.
| Characteristic | Cohort | p b | ||
|---|---|---|---|---|
| Overall, N = 390 a | Training cohort, N = 273 a | Internal test cohort, N = 117 a | ||
| Clinical manifestation | 0.178 | |||
| Other | 71 (18.2%) | 45 (16.5%) | 26 (22.2%) | |
| Abnormal uterine bleeding | 319 (81.8%) | 228 (83.5%) | 91 (77.8%) | |
| Degree of differentiation | 0.387 | |||
| Moderately or highly differentiated | 45 (11.5%) | 34 (12.5%) | 11 (9.4%) | |
| Poorly differentiated | 345 (88.5%) | 239 (87.5%) | 106 (90.6%) | |
| FIGO stage (2018) | 0.947 | |||
| IB1 | 189 (48.5%) | 132 (48.4%) | 57 (48.7%) | |
| IB2 | 201 (51.5%) | 141 (51.6%) | 60 (51.3%) | |
| BMI | 0.982 | |||
| ≤ 23.19 | 223 (57.2%) | 156 (57.1%) | 67 (57.3%) | |
| > 23.19 | 167 (42.8%) | 117 (42.9%) | 50 (42.7%) | |
| SCC | 0.480 | |||
| ≤ 2.48 | 243 (62.3%) | 167 (61.2%) | 76 (65.0%) | |
| > 2.48 | 147 (37.7%) | 106 (38.8%) | 41 (35.0%) | |
| SII | 0.224 | |||
| ≤ 477.59 | 185 (47.4%) | 135 (49.5%) | 50 (42.7%) | |
| > 477.59 | 205 (52.6%) | 138 (50.5%) | 67 (57.3%) | |
| APTT | 0.684 | |||
| ≤ 31.5 | 292 (74.9%) | 206 (75.5%) | 86 (73.5%) | |
| > 31.5 | 98 (25.1%) | 67 (24.5%) | 31 (26.5%) | |
| Fg | 0.507 | |||
| ≤ 253 | 200 (51.3%) | 143 (52.4%) | 57 (48.7%) | |
| > 253 | 190 (48.7%) | 130 (47.6%) | 60 (51.3%) | |
| ALT | 0.737 | |||
| ≤ 17 | 165 (42.3%) | 117 (42.9%) | 48 (41.0%) | |
| > 17 | 225 (57.7%) | 156 (57.1%) | 69 (59.0%) | |
| OPNI | 0.388 | |||
| ≤ 53.8 | 187 (47.9%) | 127 (46.5%) | 60 (51.3%) | |
| > 53.8 | 203 (52.1%) | 146 (53.5%) | 57 (48.7%) | |
| TC | 0.166 | |||
| ≤ 4.41 | 166 (42.6%) | 110 (40.3%) | 56 (47.9%) | |
| > 4.41 | 224 (57.4%) | 163 (59.7%) | 61 (52.1%) | |
| ALB/Fg | 0.627 | |||
| ≤ 0.17 | 137 (35.1%) | 98 (35.9%) | 39 (33.3%) | |
| > 0.17 | 253 (64.9%) | 175 (64.1%) | 78 (66.7%) | |
| HALP | 0.072 | |||
| ≤ 38.17 | 109 (27.9%) | 69 (25.3%) | 40 (34.2%) | |
| > 38.17 | 281 (72.1%) | 204 (74.7%) | 77 (65.8%) | |
n (%).
Pearson's Chi‐squared test.
3.4. Determination of the Optimal Cut‐Off Values for Continuous Variables
To avoid data leakage and overly optimistic validation performance, the optimal cut‐off values for all continuous variables were determined exclusively within the training cohort (n = 273) using receiver operating characteristic (ROC) curves and the Youden index. As shown in Figure 2, the optimal cut‐off values for tumor size (Figure 2A), BMI (Figure 2B), SCC‐Ag (Figure 2C), SII (Figure 2D), APTT (Figure 2E), Fg (Figure 2F), ALT (Figure 2G), OPNI (Figure 2H), TC (Figure 2I), ALB/Fg (Figure 2J), and HALP score (Figure 2K) to predict postoperative adjuvant therapy were determined to be 21, 23.19, 2.48, 477.59, 31.5, 253, 17, 53.8, 4.41, 0.17, and 38.17, respectively. These predefined cut‐off values were then applied to categorize variables in both the training and validation cohorts.
FIGURE 2.
Receiver operating characteristic curve combined with the Youden index was used to determine the optimal cutoff value for continuous variables exclusively in the training cohort. These pre‐defined cutoff values were subsequently applied to both the training and validation cohorts. (A: Tumor size; B: BMI; C: SCC‐Ag; D: SII; E: APTT; F: Fg; G: ALT; H: OPNI; I: TC; J: ALB/Fg; K: HALP score).
Notably, the optimal cut‐off value for tumor size was 21 mm, aligning closely with the boundary for IB1 and IB2 stages in the FIGO 2018 staging system. Given the widespread clinical use and significance of this staging system, we categorized tumor size according to FIGO 2018 in subsequent analyses: ≤ 20 mm as IB1 and > 20 mm as IB2. This ensures our predictive model meets clinical needs and effectively informs decision‐making.
3.5. LASSO Regression Variable Selection
Through LASSO regression (Figure S2), clinical manifestations (abnormal uterine bleeding), degree of differentiation (poorly differentiated), FIGO 2018 staging (IB2 stage), BMI (≤ 23.19), SCC‐Ag (> 2.48), APTT (> 31.5), TC (> 4.41), and HALP score (> 38.17) were selected, and the regression coefficients of these factors were not zero at the selected λ value (Figure S2B). As shown in Figure S3, the AUC values produced by the ROC curve analysis of the above variables were all greater than 0.5.
3.6. Univariate and Multivariate Logistic Regression
All variables selected by LASSO regression were included in the subsequent univariate and multivariate Logistic regression. As shown in Table 2, all variables were significantly related to adjuvant therapy requirement in univariate Logistic regression (p < 0.05); in multivariate Logistic regression, except for TC, the remaining variables were all significantly related to adjuvant therapy requirement (p < 0.05). TC was excluded from the final model because it lost statistical significance in multivariate logistic regression (OR = 1.59, 95% CI: 0.83–3.08, p = 0.164). Clinically, this may reflect that the prognostic effect of cholesterol is mediated through other stronger predictors already included in the model (e.g., HALP score, which encompasses nutritional status more comprehensively). To formally assess multicollinearity, we calculated the Variance Inflation Factor (VIF) for all variables in the multivariate model. All VIF values were < 2.0 (range: 1.03–1.17), well below the conventional threshold of 5.0 or 10.0, confirming the absence of multicollinearity. Thus, TC's non‐significance is attributable to its independent weak predictive effect rather than collinearity with other variables.
TABLE 2.
Univariate and multivariate logistic regression for training cohort.
| Characteristic | Univariate | Multivariate | ||||
|---|---|---|---|---|---|---|
| OR | 95% CI | p | OR | 95% CI | p | |
| Clinical manifestation | ||||||
| Other | — | — | — | — | ||
| Abnormal uterine bleeding | 5.75 | 2.34, 14.11 | < 0.001 | 5.07 | 1.74, 14.74 | 0.003 |
| Degree of differentiation | ||||||
| Moderately or highly differentiated | — | — | — | — | ||
| Poorly differentiated | 3.07 | 1.29, 7.33 | 0.011 | 5.48 | 1.81, 16.56 | 0.003 |
| FIGO stage (2018) | ||||||
| IB1 | — | — | — | — | ||
| IB2 | 8.36 | 4.75, 14.73 | < 0.001 | 6.15 | 3.13, 12.12 | < 0.001 |
| BMI | ||||||
| ≤ 23.19 | — | — | — | — | ||
| > 23.19 | 0.55 | 0.34, 0.91 | 0.020 | 0.42 | 0.22, 0.79 | 0.007 |
| SCC | ||||||
| ≤ 2.48 | — | — | — | — | ||
| > 2.48 | 5.21 | 3.08, 8.83 | < 0.001 | 3.24 | 1.64, 6.40 | < 0.001 |
| APTT | ||||||
| ≤ 31.5 | — | — | — | — | ||
| > 31.5 | 0.43 | 0.23, 0.79 | 0.006 | 0.33 | 0.15, 0.71 | 0.005 |
| TC | ||||||
| ≤ 4.41 | — | — | — | — | ||
| > 4.41 | 1.84 | 1.11, 3.04 | 0.017 | 1.59 | 0.83, 3.08 | 0.164 |
| HALP | ||||||
| ≤ 38.17 | — | — | — | — | ||
| > 38.17 | 0.44 | 0.25, 0.76 | 0.004 | 0.35 | 0.17, 0.73 | 0.005 |
Abbreviations: CI, Confidence Interval; OR, Odds Ratio.
3.7. Construction of the Predictive Nomogram
The final predictive nomogram was constructed with clinical manifestations, degree of differentiation, FIGO stage (2018), BMI, SCC‐Ag, APTT, and HALP score as predictive variables (Figure 3). According to the nomogram, degree of differentiation was considered the strongest predictor of adjuvant therapy requirement, followed by FIGO stage (2018), clinical manifestations, SCC‐Ag, APTT, HALP score, and BMI. Each predictive variable in the nomogram has a corresponding score, and the sum of the scores of all variables is the total score. By drawing a vertical line from the total score, the estimated probability of adjuvant therapy requirement was obtained.
FIGURE 3.
Nomogram with clinical manifestations, degree of differentiation, FIGO stage (2018), BMI, SCC‐Ag, APTT, and HALP score predicts the probability of adjuvant therapy requirement.
To use this nomogram clinically: First, locate the patient's value for each predictor on the corresponding horizontal axis and draw a vertical line upward to the ‘Points’ axis at the top to determine the score for that variable. Second, sum the points across all seven variables to obtain the total points. Third, draw a vertical line downward from the Total Points on the ‘Total Points’ axis to the ‘Linear Predictor’ axis, and continue to the bottom ‘Risk of adjuvant therapy requirement’ axis to read the predicted probability. For example, a patient with abnormal uterine bleeding (≈90 points), poorly differentiated tumor (≈100 points), FIGO 2018 IB2 stage (≈98 points), BMI > 23.19 (0 points), SCC‐Ag > 2.48 (≈66 points), APTT > 31.5 (0 points), and HALP score ≤ 38.17 (≈58 points) would accumulate approximately 412 total points, corresponding to a predicted adjuvant therapy requirement probability of approximately 73%. Clinicians can thus quantitatively assess individual risk and tailor treatment recommendations accordingly.
3.8. Evaluation of the Predictive Model
In the 10–fold cross–validation, the model showed a mean accuracy of 0.80 with a standard deviation of 0.04, indicating stable performance across subsets. The average recall was 0.70 and the average F1 score was 0.74, reflecting a good balance between precision and recall. As shown in Figure 4A, the AUC for the predictive model in the training and validation sets was 0.863 and 0.767, respectively, both showing good predictive performance. The calibration curves for the nomogram in the training and validation sets are shown in Figure 4B,C, and the calibration curve was relatively close to the ideal curve, indicating a good correlation between the actual and predicted incidence of adjuvant therapy requirement. The DCA (Figure 4D,E) showed that using the nomogram to predict the probability of adjuvant therapy requirement over a wide range of threshold values is beneficial, and the DCA can provide a considerable net benefit for clinical application. Specifically, in the training set (Figure 4D), the model demonstrates a significant net benefit advantage between thresholds of 0.10 and 0.80. In the validation set (Figure 4E), this advantage occurs between thresholds of 0.16 and 0.76. This indicates that the model has good clinical utility within this threshold range, providing valuable guidance for treatment decisions in practice.
FIGURE 4.
Related images to evaluate the predictive effectiveness of the nomogram. (A) The area under the receiver operating characteristic curves (AUC) for the discrimination of the model. (B and C) Calibration curves for the predicting probability of adjuvant therapy requirement in the training cohort (B) and in the internal test cohort (C). (D and E) Decision curve analysis (DCA) for the adjuvant therapy nomogram. The black line represents the assumption of no patient requiring adjuvant therapy, while the gray line assumes that all patients required adjuvant therapy. The red line corresponds to the risk nomogram. The analysis was conducted on both the training cohort (D) and the internal test cohort (E).
4. Discussion
At present, clinicians mainly rely on the FIGO 2018 staging system of cervical cancer to choose the initial treatment plan, formulate treatment strategy, and guide prognosis. However, the system can only preliminarily predict the possibility of adjuvant therapy requirement in IB1 and IB2 stage patients through tumor size, leading to a high incidence of postoperative adjuvant therapy and a large number of complications and poor health economic benefits.
How to accurately identify cervical cancer patients who are highly likely to require postoperative adjuvant therapy before surgery and perform their initial treatment shunt seems to be more conducive to reducing treatment‐related complications. In terms of identifying high‐risk pathological factors before surgery, imaging examinations have made great contributions, but there is still a need to improve their diagnostic accuracy and specificity. Magnetic resonance imaging (MRI) has excellent soft tissue contrast and is of great significance in assessing parametrial invasion and tumor size [33]; positron emission tomography/computed tomography (PET/CT) has the highest sensitivity and specificity in assessing regional lymph node status [34]. However, for high‐risk pathological factors and intermediate‐risk pathological factors that cannot be identified by imaging examinations, there is currently a lack of effective diagnostic methods, leading to poor predictive accuracy of adjuvant therapy requirement before surgery. Based on this, we defined the concept of postoperative adjuvant therapy for cervical SCC based on the Sedlis criteria [31] and first constructed a predictive model for adjuvant therapy requirement in IB stage cervical SCC with tumor size ≤ 4 cm based on preoperative clinical pathological parameters and laboratory indicators.
Our model ultimately included clinical manifestations, degree of differentiation, FIGO stage (2018), BMI, SCC‐Ag, APTT, and HALP score as seven predictive variables to construct the nomogram. The model showed good predictive performance in both the training and validation sets, which will increase the predictive efficacy of FIGO 2018 staging system for adjuvant therapy requirement in IB stage cervical SCC with tumor size ≤ 4 cm, which is closely related to the addition of more predictive variables. SCC‐Ag and degree of differentiation both affect the prognosis of cervical cancer [32, 35], the higher the value of SCC‐Ag, the lower the degree of differentiation, the worse the prognosis of the patient. Recent evidence has further elucidated the prognostic value of SCC‐Ag when integrated with inflammatory markers; a nomogram combining SCC‐Ag and neutrophil‐to‐lymphocyte ratio demonstrated superior predictive accuracy for survival outcomes in locally advanced cervical cancer [36], supporting our approach of incorporating SCC‐Ag into a multi‐parameter predictive model. In many studies, tumor size is an independent factor affecting the prognosis of IB stage cervical cancer. The larger the tumor, the more likely it is to have clinical manifestations of abnormal uterine bleeding, so clinical manifestations may be used to predict the prognosis of cervical cancer. The HALP score combines hemoglobin, albumin, lymphocytes, and platelets, which can reflect the inflammatory and nutritional status of patients with cervical cancer and may be a predictive factor for the recurrence of cervical cancer [18]. On the one hand, cytokines generated by chronic inflammation promote tumor development through multiple pathophysiological processes. On the other hand, malnutrition weakens immune function, intensifies inflammation, and increases treatment side effects in cancer patients. The prognostic relevance of systemic inflammatory indices in cervical cancer has been further validated in contemporary studies; Chen et al. demonstrated that SII independently predicted short‐term outcomes in cervical cancer patients receiving immunotherapy [37], corroborating the utility of composite inflammatory‐nutritional markers in risk stratification. Moreover, malnutrition can indicate a tumor's high metabolic activity [18]. APTT > 31.5 s emerged as a protective factor against postoperative adjuvant therapy requirement (OR = 0.33, p = 0.005). This seemingly paradoxical finding may reflect the complex interplay between coagulation and tumor biology. Prolonged APTT indicates impaired intrinsic coagulation pathway activity, which may correlate with reduced fibrinogen cross‐linking and diminished platelet‐tumor cell aggregation. In the context of gynecological malignancies, a hypercoagulable state (shortened APTT) promotes tumor cell encapsulation in fibrin‐platelet matrices, facilitating immune evasion and metastatic dissemination [38]. Conversely, patients with longer APTT may have less efficient coagulation‐driven tumor microenvironment support, resulting in lower metastatic potential and reduced need for postoperative adjuvant therapy. Additionally, prolonged APTT may be associated with lower baseline fibrinogen levels (a known pro‐tumorigenic factor), which aligns with our finding that Fg was positively associated with adjuvant therapy need in univariate analysis. This is consistent with the established concept that cancer‐associated thrombosis contributes to tumor progression, as hypercoagulability provides a scaffold for tumor cell migration and angiogenesis [38]. However, we acknowledge that APTT is influenced by multiple confounding factors (e.g., liver function, anticoagulant use), and its biological mechanism warrants prospective validation. BMI ≤ 23.19 kg/m2 was identified as a risk factor for postoperative adjuvant therapy requirement (OR = 0.42, p = 0.007), indicating that lower BMI is associated with higher adjuvant therapy requirement. This finding aligns with the ‘obesity paradox’ observed in some gynecological cancers, where higher BMI may confer survival advantages [39]. Mechanistically, lower BMI often reflects preoperative malnutrition and sarcopenia, which are associated with impaired immune function, increased postoperative complications, and higher recurrence risk [21, 24]. Malnourished patients may have attenuated anti‐tumor immunity and reduced tolerance to radical surgery, leading to more aggressive pathological features requiring adjuvant therapy. The prognostic nutritional index (PNI/OPNI) has been validated as a predictor of survival in cervical cancer, confirming the critical role of nutritional status in oncological outcomes [18]. While existing literature presents conflicting evidence regarding BMI and cervical cancer prognosis [39, 40, 41, 42], our data specifically link preoperative low BMI to increased adjuvant therapy need, likely through nutritional‐immune dysfunction rather than direct tumor biology.
Compared to existing predictive models for cervical cancer, our study offers three distinct advantages. First, unlike models predicting only lymph node metastasis [27, 28] or recurrence [18], our nomogram specifically targets postoperative adjuvant therapy requirement, directly addressing the clinical dilemma of treatment modality selection. Second, we integrated both clinical‐pathological and laboratory parameters (including the novel HALP score and APTT), whereas previous models relied predominantly on imaging or histological features alone [27, 28]. Third, our model is constructed from entirely preoperative variables, enabling true pretreatment risk stratification and therapeutic triage, in contrast to postoperative nomograms that cannot guide initial treatment decisions [43, 44]. Notably, Huang et al. [27] and Yang et al. [28] developed nomograms for pelvic lymph node metastasis prediction using hematological parameters, but these models do not address adjuvant therapy need. Our model fills this gap by providing a preoperative decision‐support tool for treatment modality selection. In clinical practice, our predictive model helps determine treatment strategies by assessing the probability of patients requiring postoperative adjuvant therapy. Patients with a probability exceeding 50% are deemed high‐risk and recommended radical radiotherapy rather than surgery. This approach reduces the incidence of postoperative adjuvant therapy, lowers complication risks, and enhances health‐economic benefits. Conversely, patients with a probability below 50% are considered low‐risk and suggested radical surgery, which is more acceptable and avoids radiotherapy‐related complications. In our cohort, 159 of 390 patients (40.77%) required post‐surgery adjuvant therapy. If our model had been used pre‐treatment to advise radical radiotherapy for those with a probability over 50%, 148 patients could have avoided postoperative adjuvant therapy, reducing its incidence from 40.77% to 2.8% (11/390). While we used a 50% probability threshold for illustrative purposes to demonstrate clinical utility, we emphasize that the optimal threshold should be determined based on institutional preferences, patient‐specific factors, and DCA‐derived net benefit curves. The nomogram provides a continuous probability estimate, allowing clinicians to select individualized thresholds according to local clinical contexts.
Reducing the complications caused by postoperative adjuvant therapy is of great significance. In terms of surgery, first, with the advancement of technology, some improved surgical methods such as nerve‐sparing radical hysterectomy can significantly improve patients' postoperative bladder function without affecting local recurrence and long‐term prognosis [11]. Second, with the improvement of cognition, people have found that the incidence of parametrial invasion in early low‐risk cervical cancer (lesions ≤ 2 cm, limited depth of stromal invasion) is low, and there is a question about whether these patients should have parametrial tissue removed. A recent randomized controlled trial showed that the 3‐year pelvic recurrence rate in low‐risk cervical cancer patients treated with simple hysterectomy was not inferior to that of radical hysterectomy, and the risk of urinary incontinence or urinary retention after simple hysterectomy was lower [45]; a recent meta‐analysis also showed similar results to the above study, that is, there was no significant difference in recurrence rate and overall survival between IA2–IB1 stage cervical cancer patients treated with simple hysterectomy or radical hysterectomy, and patients treated with simple hysterectomy had fewer postoperative complications and better surgical outcomes [46]. The above studies confirm the necessity of narrowing the indications for radical hysterectomy in early cervical cancer and reducing excessive surgery for early low‐risk cervical cancer, thereby reducing surgery‐related complications. In terms of adjuvant therapy, intensity‐modulated radiation therapy (IMRT) technology can reduce the acute toxicity of supplementary radiotherapy in cervical cancer patients after radical hysterectomy [47], and some studies have constructed predictive models for radiation enteritis in cervical cancer [43, 44], in order to identify related risk factors and carry out targeted clinical prevention and treatment, thereby reducing the side effects of radiotherapy in cervical cancer patients.
The main limitation of this study is its retrospective and single‐center design. This setup may introduce selection bias and restrict generalizability, as these are inherent flaws in such research designs that are hard to avoid. Our model performs well in the current cohort. However, the study's single‐center design, limited sample size, and potential selection bias and follow‐up limitations might restrict the model's generalizability. Therefore, the model needs to be externally validated in actual clinical practice to ensure its effectiveness and applicability in different clinical environments. Future studies should adopt a multi‐center design, expand the sample size, enhance follow‐up, and conduct external validation in diverse populations to boost the model's robustness. Another methodological consideration is the dichotomization of continuous variables (BMI, SCC‐Ag, APTT, and HALP score) for nomogram construction. While this approach enhances clinical practicality by enabling rapid bedside risk assessment, it inherently sacrifices some statistical information and assumes a uniform risk profile on either side of the cut‐off. To evaluate the impact of this simplification, we performed a sensitivity analysis retaining these four variables as continuous measures in a multivariate logistic regression model. The continuous‐variable model achieved AUCs of 0.838 (95% CI: 0.788–0.887) in the training cohort and 0.752 (95% CI: 0.660–0.843) in the validation cohort, with differences of only 0.025 and 0.015 compared with the dichotomized model (0.863 and 0.767, respectively). Calibration curves and decision curve analysis further confirmed comparable performance (Figures S4 and S5). These findings suggest that dichotomization did not substantially compromise predictive accuracy, calibration, or clinical utility, supporting its appropriateness for a clinically oriented nomogram. Nevertheless, future external validation should ideally incorporate both dichotomized and continuous formulations to confirm these observations across diverse populations. Furthermore, there are several unmeasured confounding factors in our study that may affect the accuracy of the conclusions and patients' prognosis. These include, but are not limited to, surgeons' expertise and skill levels, patients' adjuvant treatment plans, and socioeconomic status. For example, surgeons vary in their ability to achieve negative surgical margins for patients with similar surgical complexity. This can influence the need for adjuvant therapy and impact our study's conclusions. Even though surgeries in our study were performed by experienced gynecologic oncologists at our hospital, differences in surgical skill still exist. To minimize the impact of these factors, future studies should standardize surgeons' qualifications through assessment. Additionally, patients should be matched for adjuvant treatment plans and socioeconomic status to reduce their influence on study results and patient outcomes.
5. Conclusion
We successfully constructed and validated a predictive model for adjuvant therapy requirement in IB stage cervical SCC with tumor size ≤ 4 cm. Our model integrates preoperative clinical pathological characteristics and laboratory indicators, showing good predictive performance in internal validation, and can be easily applied in clinical practice, helping to reduce the need for postoperative adjuvant therapy, decrease complications, and improve health economic benefits.
Author Contributions
Kaige Pei: conceptualization, writing – original draft, writing – review and editing, methodology, data curation, software. Jiawen Zhang: conceptualization, writing – review and editing, funding acquisition, validation, project administration, resources. Dongmei Li: methodology, software, formal analysis, investigation. Mingrong Xi: supervision, visualization, validation, investigation.
Funding
The authors have nothing to report.
Ethics Statement
The study was reviewed and approved by the hospital's ethics committee, which waived the requirement for informed consent in accordance with the Declaration of Helsinki.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Figure S1: The receiver operating characteristic curves of all continuous variables during preliminary screening of predictive factors.
Figure S2: Figures related to LASSO regression. (A): Plot of binomial deviance as a function of log(λ). The left dotted vertical line indicates λ.min (minimum deviance), and the right dotted line indicates λ.1se (within 1 standard error of the minimum). (B): Plot of regression coefficients as a function of log(λ) for all 8 variables selected at λ.1se. 1—Clinical manifestation; 2—Degree of differentiation; 3—FIGO stage (2018); 4—BMI; 5—SCC; 6—SII; 7—APTT; 8—Fg; 9—ALT; 10—OPNI; 11—TC; 12—ALB/Fg; 13—HALP. Note: Although TC (variable 11) was retained by LASSO at λ.1se, it was subsequently excluded from the final nomogram after non‐significance (p = 0.164) in multivariate logistic regression.
Figure S3: The receiver operating characteristic curves for all variables that are ultimately used to build the nomogram.
Figure S4: Receiver operating characteristic (ROC) curves for the sensitivity analysis model with BMI, SCC‐Ag, APTT, and HALP score treated as continuous variables. The red line: Training cohort (AUC = 0.838, 95% CI: 0.788–0.887). The blue line: Internal validation cohort (AUC = 0.752, 95% CI: 0.660–0.843).
Figure S5: Calibration curves and decision curve analysis (DCA) for the continuous‐variable sensitivity analysis model. (A) Calibration curve for the training cohort. (B) Calibration curve for the internal validation cohort. (C) Decision curve analysis for the training cohort. (D) Decision curve analysis for the internal validation cohort. In panels A and B, the dashed blue line represents the ideal calibration, the solid red line represents the apparent calibration, and the solid green line represents the bias‐corrected calibration. In panels C and D, the red line corresponds to the continuous‐variable model, the gray line assumes all patients receive postoperative adjuvant therapy, and the black line assumes no patients receive postoperative adjuvant therapy.
Table S1: Patient demographics and baseline characteristics.
Table S2: AUCs and 95% CI for continuous variables.
Table S3: Logistic regression of categorical variables.
Acknowledgments
The authors have nothing to report.
Contributor Information
Mingrong Xi, Email: xmrjzz@126.com.
Jiawen Zhang, Email: zjw6662023@163.com.
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
References
- 1. Sung H., Ferlay J., Siegel R. L., et al., “Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries,” CA: a Cancer Journal for Clinicians 71, no. 3 (2021): 209–249. [DOI] [PubMed] [Google Scholar]
- 2. Singh D., Vignat J., Lorenzoni V., et al., “Global Estimates of Incidence and Mortality of Cervical Cancer in 2020: A Baseline Analysis of the WHO Global Cervical Cancer Elimination Initiative,” Lancet Global Health 11, no. 2 (2023): e197–e206. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Abu‐Rustum N. R., Yashar C. M., Arend R., et al., “NCCN Guidelines Insights: Cervical Cancer, Version 1.2024,” Journal of the National Comprehensive Cancer Network 21, no. 12 (2023): 1224–1233. [DOI] [PubMed] [Google Scholar]
- 4. Hwang W. Y., Kim J. H., Suh D. H., Kim K., No J. H., and Kim Y. B., “The Upper Limit of Optimal Tumor Size in Patients With FIGO 2018 Stage IB2 Cervical Cancer Undergoing Radical Hysterectomy,” International Journal of Gynecological Cancer 30, no. 7 (2020): 975–980. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Ten Eikelder M., Hinten F., Smits A., et al., “Does the New FIGO 2018 Staging System Allow Better Prognostic Differentiation in Early Stage Cervical Cancer? A Dutch Nationwide Cohort Study,” Cancers 14, no. 13 (2022): 3140. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Ryu S. Y., Deng W., Albuquerque K., et al., “Randomized Phase III Trial of Adjuvant Radiation Versus Chemoradiation in Intermediate‐Risk, Early‐Stage Cervical Cancer Following Radical Hysterectomy and Lymphadenectomy: Results From NRG Oncology/GOG‐263/KGOG 1008,” Annals of Oncology 36, no. 12 (2025): 1514–1524. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Tanaka Y., Higami S., Ishii T., Shiki Y., Ueda Y., and Kimura T., “Adjuvant Chemotherapy Versus Concurrent Chemoradiotherapy for Surgically Treated T1b Uterine Cervical Cancer: A Propensity Score Matched Analysis of the JSOG Nationwide Data,” Annals of Surgical Oncology 32, no. 3 (2025): 2206–2212. [DOI] [PubMed] [Google Scholar]
- 8. Wang Y., Yue Z. N., and Yang H., “Impact of Adjuvant Therapy After Radical Hysterectomy for Early‐Stage Cervical Cancer on Mortality and Recurrence: A Systematic Review and Meta‐Analysis,” Frontiers in Oncology 16 (2026): 1743116. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Liu P., Liang C., Lu A., et al., “Risk Factors and Long‐Term Impact of Urologic Complications During Radical Hysterectomy for Cervical Cancer in China, 2004‐2016,” Gynecologic Oncology 158, no. 2 (2020): 294–302. [DOI] [PubMed] [Google Scholar]
- 10. Bona A. F., Ferreira K. R., Carvalho R., Thuler L., and Bergmann A., “Incidence, Prevalence, and Factors Associated With Lymphedema After Treatment for Cervical Cancer: A Systematic Review,” International Journal of Gynecological Cancer 30, no. 11 (2020): 1697–1704. [DOI] [PubMed] [Google Scholar]
- 11. Yamamoto A., Kamoi S., Ikeda M., Yamada T., Yoneyama K., and Takeshita T., “Effectiveness and Long‐Term Outcomes of Nerve‐Sparing Radical Hysterectomy for Cervical Cancer,” Journal of Nippon Medical School 88, no. 5 (2021): 386–397. [DOI] [PubMed] [Google Scholar]
- 12. Wang Y., Kong W., Lv N., et al., “Incidence of Radiation Enteritis in Cervical Cancer Patients Treated With Definitive Radiotherapy Versus Adjuvant Radiotherapy,” Journal of Cancer Research and Therapeutics 14, no. 1 (2018): S120–S124. [DOI] [PubMed] [Google Scholar]
- 13. Li F., Guo H., Qiu H., et al., “Urological Complications After Radical Hysterectomy With Postoperative Radiotherapy and Radiotherapy Alone for Cervical Cancer,” Medicine (Baltimore) 97, no. 13 (2018): e0173. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Krkoska P., Kazda T., Vlazna D., and Adamova B., “Case Report: Radiation‐Induced Lumbosacral Plexopathy ‐ a Very Late Complication of Radiotherapy for Cervical Cancer,” BMC Neurology 22, no. 1 (2022): 475. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Ye X., Zhou J., Guo S., Lou P., Ma R., and Guo J., “The Undervalued Acute Leukopenia Induced by Radiotherapy in Cervical Cancer,” Current Radiopharmaceuticals 16, no. 1 (2023): 50–56. [DOI] [PubMed] [Google Scholar]
- 16. Saito M., Tanabe H., Yamauchi K., et al., “Evaluation of Non‐Hematologic Toxicity in Patients With Cervical Cancer After Type 3 Radical Hysterectomy Followed by Concurrent Chemoradiotherapy: A Retrospective Study,” Japanese Journal of Clinical Oncology 53, no. 5 (2023): 393–400. [DOI] [PubMed] [Google Scholar]
- 17. Guo J., Lv W., Wang Z., et al., “Prognostic Value of Inflammatory and Nutritional Markers for Patients With Early‐Stage Poorly‐To Moderately‐Differentiated Cervical Squamous Cell Carcinoma,” Cancer Control 30 (2023): 10732748221148913. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Jiang P., Kong W., Gong C., et al., “Predicting the Recurrence of Operable Cervical Cancer Patients Based on Hemoglobin, Albumin, Lymphocyte, and Platelet (HALP) Score and Classical Clinicopathological Parameters,” Journal of Inflammation Research 15 (2022): 5265–5281. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Lin F., Zheng R., Yu C., Su Y., Yan X., and Qu F., “Predictive Role of Serum Cholesterol and Triglycerides in Cervical Cancer Survival,” International Journal of Gynecological Cancer 31, no. 2 (2021): 171–176. [DOI] [PubMed] [Google Scholar]
- 20. Menczer J., “Preoperative Elevated Platelet Count and Thrombocytosis in Gynecologic Malignancies,” Archives of Gynecology and Obstetrics 295, no. 1 (2017): 9–15. [DOI] [PubMed] [Google Scholar]
- 21. Niu Z. and Yan B., “Prognostic and Clinicopathological Effect of the Prognostic Nutritional Index (PNI) in Patients With Cervical Cancer: A Meta‐Analysis,” Annals of Medicine 55, no. 2 (2023): 2288705. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Nomelini R. S., Mota S., and Murta E., “Absolute Band Neutrophils Count Is a Predictor of Overall Survival in Advanced Uterine Cervical Cancer,” Archives of Gynecology and Obstetrics 306, no. 5 (2022): 1697–1701. [DOI] [PubMed] [Google Scholar]
- 23. Wang H., Wang M. S., Zhou Y. H., Shi J. P., and Wang W. J., “Prognostic Values of LDH and CRP in Cervical Cancer,” Oncotargets and Therapy 13 (2020): 1255–1263. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Wang X., Xu J., Zhang H., and Qu P., “The Effect of Albumin and Hemoglobin Levels on the Prognosis of Early‐Stage Cervical Cancer: A Prospective, Single‐Center‐Based Cohort Study,” BMC Women's Health 23, no. 1 (2023): 553. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Yu J., Zheng Q., Ding X., et al., “Systematic Re‐Analysis Strategy of Serum Indices Identifies Alkaline Phosphatase as a Potential Predictive Factor for Cervical Cancer,” Oncology Letters 18, no. 3 (2019): 2356–2365. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Zhao M., Gao Z., Gu X., Yang X., Wang S., and Fu J., “Predictive Significance of Lymphocyte Level and Neutrophil‐To‐Lymphocyte Ratio Values During Radiotherapy in Cervical Cancer Treatment,” Cancer Medicine 12, no. 15 (2023): 15820–15830. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Huang Q., Li S., Chen X., et al., “A Preoperative Prediction of Lymph Node Metastasis in Early Cervical Squamous Cell Cancer With Hematologica ‐ Based Model,” Journal of Cancer 14, no. 10 (2023): 1763–1772. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Yang S., Zhao J., Zhao H., Hu Y., and Zhu H., “Development of a Nomogram for Predicting Pelvic Lymph Node Metastasis in Cervical Squamous Cell Carcinoma,” International Journal of Gynaecology and Obstetrics 160, no. 3 (2023): 1020–1027. [DOI] [PubMed] [Google Scholar]
- 29. Meng X., Song S., Li K., et al., “Application of CT in Predicting Lymph Node Metastasis in Cervical Cancer and Construction of a Preoperative Nomogram,” Scientific Reports 15, no. 1 (2025): 11674. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Yang X., Wang Y., Zhang J., et al., “A Novel Ultrasound‐Based Radiomics Model for the Preoperative Prediction of Lymph Node Metastasis in Cervical Cancer,” Ultrasound in Medicine & Biology 50, no. 12 (2024): 1793–1799. [DOI] [PubMed] [Google Scholar]
- 31. Sedlis A., Bundy B. N., Rotman M. Z., Lentz S. S., Muderspach L. I., and Zaino R. J., “A Randomized Trial of Pelvic Radiation Therapy Versus no Further Therapy in Selected Patients With Stage IB Carcinoma of the Cervix After Radical Hysterectomy and Pelvic Lymphadenectomy: A Gynecologic Oncology Group Study,” Gynecologic Oncology 73, no. 2 (1999): 177–183. [DOI] [PubMed] [Google Scholar]
- 32. Brambs C. E., Höhn A. K., Hentschel B., Fischer U., Bilek K., and Horn L. C., “The Prognostic Impact of Grading in FIGO IB and IIB Squamous Cell Cervical Carcinomas,” Geburtshilfe und Frauenheilkunde 79, no. 2 (2019): 198–204. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Merz J., Bossart M., Bamberg F., and Eisenblaetter M., “Revised FIGO Staging for Cervical Cancer ‐ A New Role for MRI,” Röfo 192, no. 10 (2020): 937–944. [DOI] [PubMed] [Google Scholar]
- 34. Choi H. J., Ju W., Myung S. K., and Kim Y., “Diagnostic Performance of Computer Tomography, Magnetic Resonance Imaging, and Positron Emission Tomography or Positron Emission Tomography/Computer Tomography for Detection of Metastatic Lymph Nodes in Patients With Cervical Cancer: Meta‐Analysis,” Cancer Science 101, no. 6 (2010): 1471–1479. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Zhu H., “Squamous Cell Carcinoma Antigen: Clinical Application and Research Status,” Diagnostics 12, no. 5 (2022): 1065. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Jin L., Cao F., Zhang Y., Dang Y., and Wang F., “Nomograms for Predicting Prognostic Value of Combined Neutrophil‐To‐Lymphocyte Ratio and SCC‐Ag in Locally Advanced Cervical Cancer,” Translational Cancer Research 13, no. 3 (2024): 1323–1335. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Chen Q., Zhai B., Li J., et al., “Systemic Immune‐Inflammatory Index Predict Short‐Term Outcome in Recurrent/Metastatic and Locally Advanced Cervical Cancer Patients Treated With PD‐1 Inhibitor,” Scientific Reports 14, no. 1 (2024): 31528. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. von Tempelhoff G. F., Nieman F., Heilmann L., and Hommel G., “Association Between Blood Rheology, Thrombosis and Cancer Survival in Patients With Gynecologic Malignancy,” Clinical Hemorheology and Microcirculation 22, no. 2 (2000): 107–130. [PubMed] [Google Scholar]
- 39. Pavone M., Goglia M., Taliento C., et al., “Obesity Paradox: Is a High Body Mass Index Positively Influencing Survival Outcomes in Gynecological Cancers? A Systematic Review and Meta‐Analysis,” International Journal of Gynecological Cancer 34, no. 8 (2024): 1253–1262. [DOI] [PubMed] [Google Scholar]
- 40. Kalliala I., Markozannes G., Gunter M. J., et al., “Obesity and Gynaecological and Obstetric Conditions: Umbrella Review of the Literature,” BMJ 359 (2017): j4511. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Frumovitz M., Jhingran A., Soliman P. T., Klopp A. H., Schmeler K. M., and Eifel P. J., “Morbid Obesity as an Independent Risk Factor for Disease‐Specific Mortality in Women With Cervical Cancer,” Obstetrics and Gynecology 124, no. 6 (2014): 1098–1104. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Leetanaporn K. and Hanprasertpong J., “Impact of Obesity on Clinical Outcomes in Patients With Early‐Stage Cervical Cancer After Radical Hysterectomy With Pelvic Node Dissection,” Oncology Research and Treatment 42, no. 11 (2019): 553–563. [DOI] [PubMed] [Google Scholar]
- 43. Ma C. Y., Zhao J., Gan G. H., et al., “Establishment of a Prediction Model for Severe Acute Radiation Enteritis Associated With Cervical Cancer Radiotherapy,” World Journal of Gastroenterology 29, no. 8 (2023): 1344–1358. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Wang J. and Hu G., “Nomogram to Predict Radiation Enteritis in Cervical Squamous Cell Carcinoma,” Cancer Management and Research 14 (2022): 3303–3311. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Plante M., Kwon J. S., Ferguson S., et al., “Simple Versus Radical Hysterectomy in Women With Low‐Risk Cervical Cancer,” New England Journal of Medicine 390, no. 9 (2024): 819–829. [DOI] [PubMed] [Google Scholar]
- 46. Taliento C., Scutiero G., Arcieri M., et al., “Simple Hysterectomy Versus Radical Hysterectomy in Early‐Stage Cervical Cancer: A Systematic Review and Meta‐Analysis,” European Journal of Surgical Oncology 50, no. 4 (2024): 108252. [DOI] [PubMed] [Google Scholar]
- 47. Contreras J., Srivastava A., Chundury A., et al., “Long‐Term Outcomes of Intensity‐Modulated Radiation Therapy (IMRT) and High Dose Rate Brachytherapy as Adjuvant Therapy After Radical Hysterectomy for Cervical Cancer,” International Journal of Gynecological Cancer 30, no. 8 (2020): 1157–1161. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Figure S1: The receiver operating characteristic curves of all continuous variables during preliminary screening of predictive factors.
Figure S2: Figures related to LASSO regression. (A): Plot of binomial deviance as a function of log(λ). The left dotted vertical line indicates λ.min (minimum deviance), and the right dotted line indicates λ.1se (within 1 standard error of the minimum). (B): Plot of regression coefficients as a function of log(λ) for all 8 variables selected at λ.1se. 1—Clinical manifestation; 2—Degree of differentiation; 3—FIGO stage (2018); 4—BMI; 5—SCC; 6—SII; 7—APTT; 8—Fg; 9—ALT; 10—OPNI; 11—TC; 12—ALB/Fg; 13—HALP. Note: Although TC (variable 11) was retained by LASSO at λ.1se, it was subsequently excluded from the final nomogram after non‐significance (p = 0.164) in multivariate logistic regression.
Figure S3: The receiver operating characteristic curves for all variables that are ultimately used to build the nomogram.
Figure S4: Receiver operating characteristic (ROC) curves for the sensitivity analysis model with BMI, SCC‐Ag, APTT, and HALP score treated as continuous variables. The red line: Training cohort (AUC = 0.838, 95% CI: 0.788–0.887). The blue line: Internal validation cohort (AUC = 0.752, 95% CI: 0.660–0.843).
Figure S5: Calibration curves and decision curve analysis (DCA) for the continuous‐variable sensitivity analysis model. (A) Calibration curve for the training cohort. (B) Calibration curve for the internal validation cohort. (C) Decision curve analysis for the training cohort. (D) Decision curve analysis for the internal validation cohort. In panels A and B, the dashed blue line represents the ideal calibration, the solid red line represents the apparent calibration, and the solid green line represents the bias‐corrected calibration. In panels C and D, the red line corresponds to the continuous‐variable model, the gray line assumes all patients receive postoperative adjuvant therapy, and the black line assumes no patients receive postoperative adjuvant therapy.
Table S1: Patient demographics and baseline characteristics.
Table S2: AUCs and 95% CI for continuous variables.
Table S3: Logistic regression of categorical variables.
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.