The working principle of PSM is to match experimental subjects with control subjects based on similarities in preexisting characteristics that may influence treatment choice. 10,11 To ensure the effectiveness of this method, the experiment and control subjects’ characteristics must be provided to determine the influence of such characteristics on treatment choice (predictor variables). Researchers use characteristic data for logistic regression analysis to estimate the association between each variable and the selected treatment. In this regression analysis, the characteristics influencing treatment choice are the independent variables, and the treatment itself is a binary dependent variable. Then, using the coefficient estimates of the regression, the predicted probability of treatment for each subject is calculated retrospectively based on the individual's specific characteristics, ranging from 0 to 1. Finally, each patient in the treatment group is matched with one or more patients in the control group based on the closest treatment probabilities and the size of the control group. There are no systemic differences between the matched groups, so these predictor variables can no longer confound group comparisons. This study is a retrospective study, and to reduce the impact of intergroup confounding effects, PSM was applied when selecting study subjects. However, PSM also has limitations. First, random controls can balance both observed and unobserved confounders, while PSM can only balance observed confounders. Therefore, residual bias is still possible. Second, an important limitation to consider is that PSM involves removing data, such as that of unmatched controls, so the sample size must be large enough to use PSM for relevant research.
Several studies have shown that compared with myomectomy, UAE results in shorter hospital stays, faster return to work, and fewer severe adverse events related to bleeding. 12,13 This study also obtained similar results: UAE resulted in less bleeding, a shorter surgery time, a shorter hospital stay, and faster physical recovery than LM. This study used four aspects of daily life as reference indicators when comparing the speed of postoperative patient recovery: time to get out of bed, time for self-care, time to complete simple household chores, and time to resume work or social activities. The results showed that UAE patients performed better in terms of all indicators (P < 0.05), directly reflecting that patients undergoing UAE surgery reported that they recovered faster and quickly returned to daily activities, housework, and work.
The Uterine Fibroid Symptom and Quality of Life (UFS-QOL) is an important tool widely used internationally to evaluate symptoms and quality of life in patients with uterine fibroids.13,14 Studies by Yeung S Y et al,15 Wei Xu et al16 and Zhou Xiaomei et al17 have confirmed that the UFS-QOL has high reliability and validity in China and is highly reflective of changes before and after treatment. Therefore, this study used the UFS-QOL to compare the symptoms and quality of life of patients with uterine fibroids treated with UAE and LM. The UFS-QOL involves 37 questions divided into two parts: the Uterine Fibroid Symptom Severity Questionnaire (UFS8, consisting of 8 questions) and the Health-Related Quality of Life Questionnaire for Uterine Fibroids (HRQL, consisting of 29 questions). The questions cover severe bleeding, menstrual conditions, urinary symptoms, psychological symptoms, sexual function, and general fatigue. Each question is divided into five levels, scored from 1 to 5 points. The raw scores are converted into a range of 0 to 100 using a dedicated UFS-QOL calculation formula.18 A higher score in the UFS8 section indicates more severe symptoms, while a higher score in the HRQL section indicates better quality of life. The results of this study based on PSM are similar to those of Manyonda et al13 and Daniels et al,14 indicating that both UAE and LM can significantly improve patients' uterine fibroid-related symptoms and significantly enhance their quality of life. However, in this study, the overall improvement in symptoms and quality of life between the two groups was similar (P > 0.05).
Studies by Zanolli et al19 suggest that fertility may not be affected for women who choose UAE, and for women who are not suitable for surgery, UAE can be offered. Serres-Cousine et al20 conducted a study on 398 patients who underwent UAE, with 148 pregnancies and 109 live births. Ludwig et al. confirmed that women can become pregnant after UAE, with many pregnancies resulting in successful deliveries. Although the actual fertility rate after UAE is still uncertain, it is close to 38.3% based on currently available results.21 In this study, three postoperative pregnancies occurred among the 66 patients in the UAE group, also confirming the fact that women can become pregnant and successfully give birth after UAE. Among these cases, one patient had infertility before surgery, suggesting that UAE may become a means for patients with fibroid-induced infertility to restore their fertility. In this study, the willingness to recommend the surgical method to other patients (1–10) was 8.8 for the UAE group and 8.5 for the LM group. This result indicates that patients who underwent either UAE or LM treatment highly approve of their surgical method and have a strong desire to recommend it to friends and family.
The strength of this study is the application of PSM to balance confounding factors between groups when selecting study subjects. It mimics some features of randomized controlled trials (RCTs) and increases the internal validity and reliability of this study. However, there are also some limitations. First, the sample size of this study is relatively small, especially after PSM. This may hinder the evaluation of differences between comparison groups, particularly potential perioperative and long-term serious complications, which are quite rare. Second, the study period was long, and there is a potential for recall bias in the long-term follow-up. In addition, some patients did not complete the questionnaire survey during the long-term follow-up. We assume that the data are missing at random, and any deviation from this assumption could lead to nonconclusive results.