Nomogram for predicting difficult total laparoscopic hysterectomy: a multi-institutional, retrospective model development and validation study

This retrospective study analyzed data from 765 patients who underwent Total Laparoscopic Hysterectomy (TLH) at two hospitals in Chongqing, China, between 1 January 2020, and 31 December 2021. Patient-specific data collected included demographic information, medical history, surgical details, and postoperative outcomes. Information about the operating surgeons was also collected. Data were analyzed using multivariate logistic regression to identify independent predictors of operative difficulty. The performance of the developed nomogram was validated both internally and externally. The dataset for this study is held and managed by the authors. Due to privacy and ethical restrictions, the data are not publicly available. Any requests for further information about the dataset and queries concerning data-sharing arrangements may be directed to Yanzhou Wang, who is presumably the corresponding author of the original study.

Highlights We create a preoperative difficulty assessment system for total hysterectomy. Surgeon factors are first introduced into the evaluation of surgical difficulty. This system can be used for developing surgical plans and also for surgeons’ training. Following caesarean section, hysterectomy is the most commonly performed gynaecological surgery, with over six million hysterectomies performed worldwide yearly 1 . Total laparoscopic hysterectomy (TLH) has emerged as the preferred approach to hysterectomy because of its advantages, such as less intraoperative blood loss, shorter hospital stays, and lower incidence of wound infection, compared to the abdominal approach 2 – 4 . However, recent research indicates a conversion rate of 0.93% from TLH to laparotomy. Moreover, the incidence of intraoperative or postoperative complications is 5.7% 5 . Therefore, performing TLH remains challenging in certain situations despite it being the preferred approach 6 . The accurate preoperative prediction of the difficulty of the surgery allows high-risk patients to be informed. Additionally, surgeons can better schedule the surgery and prepare a competent team to provide increased medical support for high-risk patients. Moreover, surgeons should be prepared for potential complications 7 – 10 . Several studies have evaluated the difficulties associated with TLH 9 , 11 . However, these studies primarily focused on the incidence of converting to laparotomy, postoperative complications, and readmission rates 12 , 13 . Using retrospective studies, researchers attempted to identify high-risk factors contributing to these occurrences. The literature reported several common independent risk factors, including uterine weight and width, surgical history related to pelvic and abdominal adhesions, BMI, and uterine artery position 11 , 14 , 15 . Since these factors can provide insights into the difficulty of the surgery, they may serve as indicators for surgeons to proceed with caution. Matsuo and colleagues conducted a further analysis of these risk factors. They found that the conversion to laparotomy was 1.1% among patients without risk factors. However, this increased to 21.7% among patients with two risk factors. When three risk factors were present, the rate of conversion to laparotomy was 50% 13 . However, the contribution of each risk factor to the surgical difficulty varied. Thus, making a personalized risk prediction is difficult. Preoperative scoring systems are currently used in specific surgical fields, including hepatobiliary surgery. Ewen and colleagues developed a scoring system for laparoscopic cholecystectomy that utilized the preoperative patient characteristics to predict surgical complexity. The system demonstrated positive predictive values of 90% and 88% for uncomplicated and complex cases, respectively 16 – 18 . However, effective tools for predicting the operative difficulty in TLH are insufficient. Therefore, a predictive scoring system that accurately identifies high-risk patients and predicts difficult surgeries should be established to reduce surgical complications and enhance surgical efficiency. This study was designed to identify high-risk factors influencing operative difficulty and to develop an internally and externally validated preoperative scoring system for personalized and precise assessment of TLH.

Y.C.: conceptualization, formal analysis, investigation, visualization, writing—original draft. J.J.: investigation, writing—original draft. M.H.: investigation, writing—original draft. K.Z.: investigation. S.T.: writing—review and editing. L.D.: validation, writing—review and editing. Y.W.: conceptualization, funding acquisition, resources, supervision, writing—review and editing.

None.

This study involves the use of patient data that has been anonymized and does not contain any personally identifiable information. The data used are publicly available and/or the use of such data for research purposes does not require specific consent as per the guidelines of the Institutional Review Boards of Southwest Hospital, Army Medical University, and the 958th Army Hospital of the Chinese People's Liberation Army. Therefore, the requirement for patient consent was waived.

This research was approved by the Institutional Review Boards of Southwest Hospital, Army Medical University and the 958th Army Hospital of the Chinese People's Liberation Army (approval numbers: (B) KY2023068 and IRB20231k-2 respectively). All patient records used in this study were anonymized to ensure patient confidentiality. During the data collection and analysis, we strictly adhered to applicable data protection and privacy regulations. No procedures that could potentially cause harm or inconvenience were performed on the participants as part of this research.

This was a retrospective study involving patients who underwent TLH at two tertiary hospitals (The First Affiliated Hospital of Army Medical University (Southwest Hospital) and The 958th Army Hospital of the Chinese People's Liberation Army (958th Hospital))in Chongqing, China. Both hospitals have proficient laparoscopic technology, and almost all hysterectomy surgeries are performed laparoscopically, unless the patient is unable to tolerate laparoscopy or strongly prefers open abdominal surgery. The study protocol was approved by the Ethics Committees of Southwest Hospital (approval number: (B) KY2023068) and 958th Hospital (approval number: IRB20231k-2), and the requirement for informed consent was waived. This study was retrospectively registered in the Chinese Clinical Trial Registry (ChiCTR). The registration number was ChiCTR2300075298. The registry entry can be accessed at [ https://www.chictr.org.cn/bin/project/edit?pid=205440 ]. This work has been reported in line with the Standards for the Reporting of Diagnostic Accuracy Studies (STARD) criteria and the Strengthening the Reporting of Cohort Studies in Surgery (STROCSS) criteria 19 , 20 . We calculated the minimum sample size required for developing the clinical prediction model to be 329, based on the four-step approach proposed by Riley et al . 21 . This method takes into account the number of predictors, the anticipated event rate, and performance metrics of the model to ensure its accuracy and to prevent overfitting. This retrospective study included 663 patients who were hospitalized at Southwest Hospital between 1 January 2020, and 31 December 2021. The patients were divided into training and internal validation sets using a 7:3 ratio. Additionally, 102 patients hospitalized at 958th Hospital between 1 January 2020, and 31 December 2021, were included in the external validation dataset. The inclusion criterion was patients underwent TLH for benign uterine disease. Patients with malignant tumours, those who underwent TLH with concurrent surgery, and those with incomplete medical records were excluded (Fig. 1 ). Flowchart of patients who underwent total laparoscopic hysterectomy included in the developmental dataset and external test dataset. TLH, Total laparoscopic hysterectomy. In developing our prediction model, we reviewed literature, 10–11,13 consulted experts, and leveraged clinical experience to select potential predictive factors from electronic health records for analysis. This included age, pregnancy status, parity, height, weight, BMI, history of pelvic surgery, the presence of anaemia, dysmenorrhoea, uterine tenderness and diagnosis. Ultrasound data were collected, encompassing measurements of uterine length, width and anteroposterior diameter, as well as the aggregate of the diameters of the three largest fibroids and the location of the fibroids. The uterine weight was calculated using the method proposed by Q. Sheng et al . 22 , with the formula: uterine weight = 1 g/cm 3 * 0.52 × Length × Width × Anteroposterior diameter. Other collected information included surgical procedure, operation time, blood loss and complications. Furthermore, information about the operating surgeons, including their sex, years of experience, and the number of hysterectomies performed in the study period, was collected. The patient is positioned in the lithotomy position, and a uterine manipulator is placed based on the vaginal and cervical measurements. A 10-mm trocar is inserted at the umbilicus for the laparoscope, and three other 5-mm trocars are inserted separately at 2 cm above the anterior superior iliac spine on both sides and at the midpoint between the anterior superior iliac spine and the umbilicus for the operative instruments. At the onset of the surgery, the fallopian tube ligaments, the utero-ovarian ligaments, as well as the round ligaments are coagulated and cut using bipolar coagulator and scissors. The anterior and posterior leaves of the broad ligament are then separated, identifying the peritoneal fold between the bladder and the uterus and dissected forward to detach the bladder from the lower segment of the uterus. The uterine vessels are skeletonized, and the ascending uterine vessels are coagulated at the level of the internal os of the cervix using bipolar forceps. Following this, a circumferential incision is made in the vaginal wall along the impression formed at the vaginal apex by the uterine manipulator. If the uterus is moderately sized, it is directly removed through the vagina. For a too-large uterus, vaginal morcellation is performed for removal. Finally, the vaginal cuff is sutured to complete the surgery. A difficult TLH was defined using the following criteria: (1) conversion to open abdominal surgery; (2) surgical time exceeding the 75th percentile (≥ 130 min); (3) intraoperative blood loss greater than 500 ml; and (4) injury to adjacent organs, such as the bladder or rectum. Univariate logistic regression was initially utilized to evaluate the association between each independent variable and difficult surgery. Significant factors identified from this univariate analysis ( P < 0.1) were then considered for inclusion in a multivariate logistic regression model. A backward selection approach was applied, using the Akaike Information Criterion (AIC) for model selection, to identify the factors for the final model, which were included in the development of the nomogram. Predictor lines were drawn upwards to assign points based on the nomogram. The sum of these points was located on the “Total Points” axis. The line, projecting downward to the bottom scales, indicated the possibility of a difficult TLH. The model was evaluated in terms of its sensitivity, specificity, accuracy, positive and negative predictive values, and area under the receiver operating characteristic curve (AUC). Additionally, a calibration curve was used to assess the predictive accuracy of the model, and a decision curve analysis (DCA) was used to evaluate the net benefit of the model for patients. Statistical analyses were performed using R version 4.0.3 software, using the rms, pROC, ggplot2, and DCA packages. For continuous data, normally distributed variables were presented as mean ± standard deviation, while skewed variables were presented as median (M) and interquartile range (P25–P75). Count data were presented as percentages (%). All statistical tests were two-sided, and a P value less than 0.05 was considered significant.

Based on the aforementioned definition, 169 (25.49%) and 37 (36.27%) cases were classified as complex operations from Southwest Hospital and 958th Hospital, respectively. Demographic comparisons between the two hospitals showed no significant differences in disease-related factors, including a history of pelvic surgery, myoma position, the sum of myoma diameters, and uterus size. However, significant differences were observed in the surgeons’ characteristics. In particular, surgeons from Southwest Hospital had more years of practice and performed more hysterectomies than those from 958th Hospital. The two hospitals also had significant differences in terms of the surgery-related data. Higher rates of complications and conversion to laparotomy were observed in 958th Hospital. The patients in the 958th Hospital dataset had significantly higher intraoperative bleeding volumes and longer operative times. All the characteristics of the patients and surgeons are presented in Table 1 and Table 2 . Patients’ characteristics. Data are reported as N (%) or median (interquartile range), with p values from χ 2 or Mann–Whitney tests, respectively. P values are significant at P < 0.05. HSIL, high-grade squamous intraepithelial lesion; LBA, laparoscopic bilateral adnexectomy; LBS, laparoscopic bilateral salpingectomy; TLH, total laparoscopic hysterectomy. Surgeons’ characteristics Calculated from December 2022. Based on the univariate analysis of the training dataset, the following factors predicted a difficult operation: the surgeon's characteristics, particularly the number of years in practice and annual hysterectomy volume; the patients’ characteristics, including the presence of myoma and adenomyosis; uterine size in terms of the length, width, anteroposterior diameter and weight; the sum of the diameters of the three largest myomas; the position of the myoma; and the presence of anaemia. Based on the multivariate analysis, the significant independent predictors of difficult operations were the number of years in practice and annual hysterectomy volume of the surgeon, as well as the uterine weight, history of pelvic surgery, and presence of adenomyosis (Table 3 ). Univariate and multivariate analyses for factors associated with difficult surgery. HSIL, high-grade squamous intraepithelial lesion; OR, odds ratio. Based on the multifactorial results of the logistic regression analysis, a nomogram was established to predict the likelihood of a difficult operation (Fig. 2 A). Each risk factor was assigned a score. The total score was calculated by adding the individual scores and locating the sum on the total-point scale axis. A vertical line was drawn downward from this point to determine the probability of a difficult operation. Nomogram model predicting the likelihood of a difficult surgery. (A) The nomogram is used by summing all points identified on the scale for each variable. The total points projected on the bottom scales indicate the possibility of a difficult operation for patients undergoing TLH. ROC curves of the training dataset (the cutoff value is 0.306) (B) and validation dataset (C). Calibration curves predict the likelihood of a difficult surgery in the training dataset (D) and validation dataset (E). Decision curve analysis for the training dataset (F) and the validation dataset (G). ROC, receiver operating characteristic; TLH, total laparoscopic hysterectomy. The nomogram demonstrated a robust predictive performance in both the training and internal validation datasets. The AUCs were 0.827 (95% CI, 0.783–0.872) and 0.793 (95% CI, 0.714–0.872), respectively (Fig. 2 B, C). The calibration curves for the complex cases showed good agreement (Fig. 2 D, E). Additionally, a DCA of the nomogram was conducted (Fig. 2 F, G), and it demonstrated that a model-based decision provided a more significant net benefit, compared to no treatment or an all-treatment approach, for predicted probability thresholds ranging from 0 to 80%. External validation was performed using a dataset of 102 surgical cases at the 958th Hospital. Despite the differences in patient characteristics between the two hospitals (Table 1 ), the nomogram maintained a good predictive accuracy with an AUC of 0.756 (95% CI, 0.658–0.854) (Fig. 3 A). The calibration curve also had a good performance (Fig. 3 B), and the DCA indicated that patients achieved a high net benefit for the predicted probability thresholds between 0 and 80% (Fig. 3 C). External validation for predictive accuracy of the nomogram. (A) ROC curves for the external validation dataset. (B) Calibration curves for predicting difficult surgery in the external validation dataset. (C) Decision curve analysis for the external validation dataset. ROC, receiver operating characteristic; TLH, total laparoscopic hysterectomy. To optimize the predictive performance of our model, we identified the optimal balance point on the ROC curve as the cutoff value. Cases with a difficulty possibility greater than 0.3 were defined as difficult surgeries (Fig. 2 B). The sensitivity, specificity, accuracy, positive predictive value, and negative predictive value of the three datasets are presented in Table 4 . Model performance. Cutoff value = 0.3. Data are in percentages (%) (95% CI).

ChiCTR2300075298.

The Guarantors, who accept full responsibility for the work, are Authors Yanzhou Wang and Yin Chen.

This study created a preoperative nomogram that predicted the operative difficulty of TLH and was validated internally and externally. Notably, the nomogram includes surgeon characteristics, making it applicable to surgeons with varied experience levels. The ability to predict the difficulty of the operation improves the preoperative planning process, allows more appropriate patient education, and predicts adverse outcomes, including the likelihood of converting to a laparotomy. Additionally, a nomogram can assist surgeons in determining the necessity for a specialist or senior surgeon in cases that are deemed difficult. Furthermore, it can be used to optimize surgical training at different levels because cases may be assigned to match the trainee's level.

This was the first study to establish a nomogram, that predicted the difficulty of TLH in patients with benign gynaecological diseases. The results of this study demonstrated that several factors contributed to the difficulty of TLH. These included patient characteristics, such as uterine weight, presence of adenomyosis, and history of pelvic surgery, as well as surgeon factors, such as years of experience and annual hysterectomy volume. Among the risk factors evaluated in this study, uterine weight had the most significant impact on surgical difficulty. Additionally, surgeons with a larger annual hysterectomy volume and more years of experience tended to have less difficulty performing TLH. This factor was rarely considered in previous analyses of risk factors for difficult operations. Furthermore, the nomogram was validated externally using data from a second hospital to confirm its reliability. Minimally invasive surgery (MIS) has become the primary surgical method for total hysterectomy. The use of MIS techniques offers numerous advantages in reducing blood loss and postoperative complications and enhancing recovery after surgery. However, every surgical procedure is inherently variable, and certain cases may present difficulties for surgeons, potentially affecting the successful completion of laparoscopic hysterectomy 23 . Interestingly, defining what constitutes a “difficult surgery” itself is challenging. Currently, most studies identify a surgery as difficult based on adverse events occurring during or after the procedure and utilize certain outcome measures as surrogate markers for defining “difficult surgery.” In studies on the difficulty of laparoscopic total hysterectomy and laparoscopic colon resection, researchers have unanimously chosen surgical time, blood loss, Dindo–Clavien II–III grade complications, conversion to open surgery, and length of hospital stay as surrogate markers 24 , 25 . Our study also drew from the definition of “surgical difficulty” in other surgical fields and defined surgical difficulty based on four perioperative indicators: surgical time exceeding the 75th percentile (surgical time ≥130 min), total blood loss greater than 500 ml, conversion to open surgery, and occurrence of adjacent organ injury. As the length of hospital stays in our data was relatively short and showed no difference, we excluded it from our study. Previous studies indicated that uterine dimensions, including the width, length, anteroposterior diameter, and weight, were significant in determining the difficulty of TLH 14 , 26 , 27 . In the present study, the predictive values of these factors were investigated, and only the uterine weight was identified as a significant risk factor based on the multiple regression analysis. Specifically, the uterine weight was directly proportional to the difficulty score. This finding was consistent with previous reports, which demonstrated that uterine weight was significantly correlated with intraoperative bleeding, operation time, conversion rate, and complications. The uterus has an elongated ellipsoidal shape. Since its length, width, and height are interrelated, the complexity of the operation is increased. Calculating the weight of the uterus, based on these three dimensions, provides a more comprehensive assessment of its size 22 , 28 . A larger uterus is often associated with a more abundant blood supply and more significant variations in the position of the blood vessels. These ultimately increase the risk of bleeding and require more complex management. Additionally, a larger uterus can compress the neighbouring organs, increasing the rate of unintended damage 9 , 26 , 29 . Similarly with previous reports, the multivariate logistic regression analysis of the present study revealed that a history of pelvic surgery and adenomyosis significantly increased the probability of difficult surgery 30 . Previous studies suggested that pelvic adhesions were significant risk factors for surgical complications and conversion to laparotomy. Additionally, a history of pelvic surgery, particularly caesarean section, which is prevalent in China, can result in postoperative adhesions between the bladder and cervix, further complicating the separation of the bladder from the uterus 31 , 32 . In the present study, both patients who experienced bladder injuries had previously undergone two caesarean sections. Adenomyosis is frequently observed as a contributing factor to developing adhesions and complications in TLH. Adenomyosis is a predictive factor for posterior cul-de-sac obliteration and deep-infiltrating endometriosis, which can lead to significant surgical morbidity 33 . This study represents the inaugural attempt to integrate surgeon demographics, encompassing variables such as gender, years of experience, and annual surgical volume, within a preoperative scoring framework aimed at assessing operative complexity for hysterectomy procedures. It also identified the years of experience and annual surgical volume as risk factors, associated with surgical difficulty. Arora et al. 34 . compared surgical experience and skill using a high-fidelity TLH model and found that gynaecology surgery specialists, averaging 8.9 years in practice, scored higher than the residents and fellows. Mehta et al. 35 . conducted a systematic review of 14 peer-reviewed international studies on gynaecology, gynaecologic oncology, and urogynecological patients, undergoing hysterectomy, and found that a dataset of low-volume surgeons (≤ 12 hysterectomies) had an increased rate of complications. In our research, we have also observed analogous outcomes. Junior doctors emerge as a significant contributory factor in a challenging context wherein encountering additional challenging factors may augment the likelihood of surgical complications or necessitate conversion to laparotomy. However, our research has certain limitations. Firstly, it is a retrospective study, which could introduce potential data bias. Secondly, our study has primarily concentrated on patients residing in southern China, thereby limiting the generalizability of our research findings to populations in other countries and regions. Factors such as patients’ ethnicity, BMI, response to surgical stimulation, as well as disparities in medical technology and healthcare policies across different countries and regions, have the potential to influence the outcomes. In our forthcoming studies, we intend to address these factors and carry out more comprehensive research.

Not commissioned, externally peer-reviewed.

The authors confirm that they have no conflicts of interest.