Introduction

Cervical cancer stands as the fourth leading cause of cancer among women globally, with around 604 000 new cases and 342 000 deaths in 2020.1 High-risk human papillomaviruses (HPV) are the primary cause, with smoking, immunosuppression and long-term contraceptive use also contributing.2–6 Efforts in early detection and HPV vaccination have reduced its impact in developed countries.7 Yet, in less developed regions, the disease’s prevalence remains high due to limited access to preventive measures.8 9

Surgery is crucial for treating early-stage cervical cancer. The 2020 guidelines from the National Comprehensive Cancer Network (NCCN) recommend radical hysterectomy with pelvic lymphadenectomy up to stage IIB.10 Such surgeries, however, disrupt the nerves controlling bladder function, leading to postoperative urinary retention (POUR).11 12 POUR is defined by the inability to void after surgery, which is commonly associated with partial detrusor denervation and increased myogenic tonicity of the detrusor muscle due to surgical trauma and prolonged catheter drainage.13

Also, de-afferentiation of the bladder wall and Fowler’s syndrome have been implicated in the development of urinary retention after hysterectomy.14 Urodynamic studies suggest that if the micturition reflex is ablated during surgery, a reduction in urethral pressure is necessary for balanced bladder function, indicating that nerve supply to the urethra is affected.15 Partial or complete denervation of the bladder and proximal urethra due to surgical radicality is another major cause of urinary retention, with functional obstruction also playing a role in cases of stress-induced urinary incontinence surgery.16

POUR occurs in 5%–70% of cervical cancer patients, with 29% following radical hysterectomy, according to earlier reports.17 18 This suggests a unique risk associated with cervical cancer surgeries, leading to complications such as urinary tract infections (UTIs), bladder overdistension, catheter-related complications, autonomic dysregulation, increased morbidity, prolonged hospital stays, additional posthospitalisation care, increased hospital costs and dissatisfaction with surgery outcomes.19–22 The condition severely affects patients' quality of life, often necessitating long-term catheter use and risking permanent bladder damage.23

Besides advancements in novel surgical techniques, interventions such as low-frequency electrical stimulation, oral Misoprostol and optimising the duration of postoperative catheterisation can be effective measures to reduce the risk of urinary retention following radical hysterectomy.24–26 Despite known risk factors such as older age, longer surgery duration, larger volumes of intraoperative fluids, POUR management in cervical cancer remains challenging.27 High rates of retention and reliance on catheters indicate a need for improved care strategies.

This study aims to identify factors linked to POUR and develop a predictive model using data collected from patients undergoing cervical cancer surgery. By validating this model clinically, the results may aid in early POUR detection and prevention and ultimately improve patient outcomes and quality of life after cervical cancer surgery. Our work may contribute to advancing care for cervical cancer surgery patients, reducing postoperative complications, and enhancing recovery processes.

ResultsDemographic characteristics

A total of 1159 patients were initially enrolled, with 58 patients excluded (5.0%). The eventually included 1101 patients were predominantly stage I (63.4%) and II (30.6%) in terms of tumour staging, with a small fraction of the cohort having undergone radio/chemotherapy before surgery (14.3%). Half of them were aged 40–54 years (51.3%), and the vast majority are married (87.1%).

Most patients had a normal BMI (68.9%), though a significant proportion was obese (17.5%). Over one-third (34.3%) had a history of urinary system disease, with a significant proportion experienced ureteral adhesion (62.3%). The majority healed well (primary healing, 82.3%). A large proportion used an analgesia pump (62.8%). In contrast, few engaged in bladder training (10.9%).

About half were without anxiety and depression, 52.3% and 54.3%, respectively (table 1).

Table 1

Demographic and clinical characteristics of patients (n=1101)

Influencing factors of postoperative urinary retention (POUR) in cervical cancer patientsUnivariate analysis

Out of 1101 patients undergoing cervical cancer surgery, 393 experienced POUR (35.69%). When comparing possible factors influencing occurrence of POUR between the POUR and the non-POUR group, univariate analysis results showed that the duration of surgery, intraoperative bleeding, presence of diabetes, hypertension, ureteral adhesion, wound healing classification, preoperative radio/chemotherapy, category of BMI, history of urinary diseases, history of caesarean section, postoperative urinary infection and the use of analgesia pumps were significantly higher in the POUR group, with all differences being statistically significant (p<0.05) (table 2).

Table 2

Significant variables identified in univariate analysis (p<0.05, n=1101)

Multivariate logistic regression analysis

The statistically significant influencing factors from the univariate analysis, along with other clinically significant variables (anxiety category, depression category, pain numerical rating score), were subjected to stepwise Logistic regression model construction, with entry at α=0.05 and removal at α=0.1.

The results indicated that diabetes, wound healing classification, presurgery radio/chemotherapy, postoperative urinary infection, use of analgesia pumps and pain numerical rating score were founded to influence occurrence of POUR in cervical cancer patients (p<0.05) (table 3).

Table 3

Logistic multivariable regression analysis of influencing factors for POUR in cervical cancer patients

Predictive modelling

The predictive probability value was used as the test variable, and the occurrence of POUR was the state variable, to evaluate the discrimination ability of this prediction model using the ROC curve. The results showed that the area under the ROC curve (AUC) for the predictive model was 0.897, with a 95% CI of 0.877 to 0.916 (p<0.001). The model’s specificity at the optimal threshold was 0.814, and sensitivity 0.847 (figure 1).

Figure 1

ROC curve of predictive model for POUR in cervical cancer patients. The AUC for the predictive model was 0.897, with a 95% CI of 0.877 to 0.916 (p<0.001). The model’s specificity at the optimal threshold was 0.814, and sensitivity 0.847. AUC, area under the curve; POUR, postoperative urinary retention; ROC, receiver operator characteristic.

The model’s calibration was assessed using the Hosmer-Lemeshow goodness-of-fit test, which assesses whether the observed event rates match expected event rates in subgroups of the model population by dividing the data into deciles based on predicted probabilities and comparing the observed and expected counts of events in each decile using a χ2 test. A significant p value, typically<0.05, suggests that the model does not adequately fit the data. In our case, it showed a Hosmer-Lemeshow χ2 of 30.239 (p<0.001) (figure 2).

Figure 2

Hosmer-Lemeshow goodness-of-fit test of predictive model of POUR in cervical cancer patients. Despite Hosmer-Lemeshow χ2 of 30.239 (p<0.001), this did not affect the validity of our logistic regression model, which was statistically significant with an R2 value of 0.583. POUR, postoperative urinary retention.

Notably, despite the poor fit, this did not affect the validity of our logistic regression model, which was statistically significant with an R2 value of 0.583. Additionally, with an AUC of 0.897 (p<0.001), it further indicated that our regression model was usable.

Validation of prediction model

We performed an external validation of the prediction model using a cohort of 205 patients undergoing cervical cancer surgery at our hospital between 1 August 2023 and 31 October 2023 using the same criteria (table 4).

Table 4

External validation of prediction model (n=205)

According to the external validation, the sensitivity of the prediction model was 0.591 and specificity 0.747, demonstrating good prediction power. In addition, we calculated the Positive Predictive Value (PPV) and Negative Predictive Value (NPV) to further assess the performance of our predictive model. The PPV, which indicated the probability that patients identified as at risk for POUR actually experience the condition, was approximately 55.4%. The NPV, which indicated the probability that patients not identified as at risk truly did not develop POUR, was approximately 72.7%.

Discussion

In our study, we identified diabetes, wound healing classification, preoperative radio/chemotherapy, postoperative urinary infection, analgesia pump use and pain numerical rating score as significant risk factors for POUR in cervical cancer patients.

Patients with diabetes are at a higher risk of developing POUR: diabetes mellitus is generally considered a risk factor for POUR following various surgical procedures.31 32 This is consistent with our finding that diabetic patients are more likely to develop POUR than their non-diabetic counterparts, potentially due to its effects on blood sugar levels, disruption of micturition pathways and associations with conditions like gut dysbiosis and cognitive impairment that may indirectly affect urinary function.33 The complexity of diabetes combined with urinary retention presents significant treatment challenges, warranting active and effective dietary and insulin therapies.

Patients with poor postoperative wound healing are prone to urinary retention: neurogenic bladder dysfunction has long been considered a major cause of POUR in cervical cancer patients.17 34 We found that patients with secondary and tertiary wound healing are more likely to develop POUR than those with primary wound healing, which is consistent with previous reports. This may be due to a variety of factors including surgery type, anaesthesia, comorbidities and pain, leading to complications, such as bladder overdistension, detrusor damage, increased risk of infection and delayed hospital discharge.17 34 35 Previous studies have suggested a number of measures, which may improve postoperative healing, such as a combination of nutritional support, preoperative interventions, perioperative management and postoperative pain management with regional anaesthetic techniques and local infiltration of medications.36–39

Patients who underwent preoperative radio/chemotherapy are at increased risk of POUR: preoperative radio/chemotherapy is not conventionally considered a risk factor of POUR. However, our analysis identified it as a possible influencing factor where patients who had preoperative radio/chemotherapy are more likely to develop POUR than those who did not. This association and reduced likelihood of POUR in cervical cancer patients could be explained by several hypotheses, that is, reduced tumour size leading to less extensive surgical dissections, minimised disturbance to the pelvic floor and bladder during surgery, alterations in the inflammatory and healing responses due to radio/chemotherapy, differences in clinical characteristics and surgical approaches for patients receiving preoperative treatments and finally potential biases in monitoring and reporting POUR. This suggests a complex interplay between preoperative treatments and postoperative outcomes, which requires a comprehensive understanding of the factors influencing POUR risk and future research.

Patients with postoperative urinary infections are prone to POUR: while POUR is a well-documented cause of UTIs postsurgery, our findings indicate that patients with postoperative UTIs are more likely to develop POUR,40 which suggests a complex, possibly bidirectional relationship between these conditions. This relationship could be attributed to several factors, such as the inflammatory responses from UTIs potentially causing functional obstructions or decreased bladder compliance, alterations in neural bladder control leading to dysfunctional voiding, pain and discomfort associated with UTIs resulting in voluntary urinary retention, direct impairment of bladder detrusor muscle function and psychological stress influencing bladder control. This reverse sequence from the more common pathway where POUR leads to UTIs is less explored in existing literature, which calls for further research to elucidate underlying mechanisms and improve postoperative care.

However, this does not diminish the importance to address postoperative UTIs in clinical practice as it remains critical for patient recovery and overall outcomes. Addressing UTIs promptly can prevent the escalation of infections into more severe complications and mitigate their potential role in precipitating or exacerbating POUR. Effective management strategies include early detection through vigilant monitoring of symptoms and urinalysis, appropriate antibiotic therapy guided by culture and sensitivity results and strategies to promote urinary tract health, such as encouraging adequate hydration and bladder training exercises.41 42

Patients using analgesia pumps are more prone to POUR: using a patient-controlled analgesia (PCA) pump postoperatively is reported to increase the risk of POUR in patients undergoing elective spinal surgery,20 which is consistent with our finding that cervical cancer patients using analgesia pumps after surgery were more likely to develop POUR than those who did not. The use of PCA pumps may delay bowel motility, leading to adynamic ileus, which could contribute to the development of POUR by affecting the autonomic nervous system that controls bladder function.43 Despite the benefits of pain pumps in reducing postoperative pain and potentially lowering the risk of complications such as deep vein thrombosis, the increased POUR risk suggests a need for careful monitoring of bladder function when these devices are employed.44

Patients with high pain scores are prone to POUR: there is an established association between pain and the occurrence of POUR. Higher postoperative pain scores are linked to an increased risk of POUR, and pain management strategies are being explored as a means to prevent this complication.45 46 According to our results, higher pain scoring suggests greater likelihood of POUR. Effective pain management strategies, including the use of such medications as perioperative non-steroidal anti-inflammatory drugs and glycopyrrolate, may help minimise this risk.47 Interestingly, the use of PCA pumps, as discussed above, may paradoxically elevate risk of POUR, which could be attributed to the opioids often used in PCA pumps, medications known to affect bladder function by increasing bladder sphincter tone and reducing the sensation of bladder fullness, leading to urinary retention.48 It is essential, therefore, to balance the need for effective pain control with the potential for contributing to POUR when utilising opioid-based PCA for postoperative analgesia. Furthermore, multimodal pain management strategies that incorporate non-opioid analgesics and regional anaesthesia techniques can offer significant advantages.

The POUR predictive model for cervical cancer patients developed and validated in our study demonstrates good predictive power, with an AUC of 0.897 (95% CI, 0.877 to 0.916, p<0.001). The sensitivity of the model at the optimal threshold was 0.591, with specificity being 0.747. Existing studies have reported a variety of prediction models for different outcomes in cervical cancer patients, but none exists for POUR.49–51 Its excellent performance and predictive power supports potential clinical utilisation. Future efforts could focus on improving its sensitivity without substantially compromising specificity by incorporating additional predictive factors, such as the extent of surgical dissection, intraoperative nerve damage, use of certain anaesthetic agents and individual patient characteristics.

The inclusion of PPV and NPV in our model’s evaluation presents a more in-depth view of its diagnostic performance. While the specificity and sensitivity provide a measure of the model’s overall accuracy, the PPV and NPV offer direct clinical relevance by demonstrating the model’s effectiveness in a real-world setting. This is particularly important in managing POUR where the consequences of false positives and false negatives carry significant clinical implications. The PPV of 55.4% indicates the model’s effectiveness in predicting actual cases of POUR, guiding necessary interventions. Similarly, an NPV of 72.7% assures clinicians that patients not flagged by the model are truly at low risk, minimising unnecessary treatments or interventions. Future enhancements of the model should focus on optimising the values to improve the model’s utility to ensure that it can reliably support clinical decisions with minimal risk to patient care.

It is worth noting that the 2 week period in the current study was specifically chosen for cervical cancer patients based on the acute recovery timeline typically observed in this group. While appropriate for capturing immediate postsurgical complications in cervical cancer, this timeframe may not be suitable for other populations such as those with non-oncological urogynecological conditions, where longer-term complications might occur. Given the potential for different recovery dynamics across patient groups, it would be beneficial for future research to explore different durations of monitoring POUR to better understand its impact over an extended period.