Epidemiological characteristics of suspected adenomyosis in the Chinese physical examination population: a nested case-control study

Adenomyosis (AM) is defined as ectopic endometrial glands and stroma within the myometrium, causing hyperplasia and enlargement of the myometrium, resulting in an enlarged uterus. 1 AM is a common disease and its prevalence diagnosed by transvaginal ultrasound is 20.9%–21.8%, 2 3 usually affecting women over the age of 40. 4 Symptoms related to AM include abnormal uterine bleeding, pelvic pain (dysmenorrhoea, chronic pelvic pain, dyspareunia) and impaired reproduction, 5 seriously impacting patient quality of life. The average time from the first occurrence of AM-related symptoms to diagnosis is 5.7 years, and 41.9% of patients report severe/very severe AM. 6 Therefore, although it takes substantial time for the disease to progress from symptomatic status to disease diagnosis, patients have little chance to delay the deterioration of disease. As hysterectomy is the gold standard for diagnosis of the disease, most studies have been based on women who had a hysterectomy. 1 7–11 However, the selection bias of women with hysterectomy may result in an overestimate on the prevalence of the disease. 12 Alternatively, population-based cohort studies on AM that used hysterectomy as the diagnostic criteria may underestimate the incidence rate, 13 due to the small population of hysterectomy for AM. With the development of imaging techniques, ultrasound has been able to clearly define the characteristics of AM, 14 with both high sensitivity and specificity. 15 Yu et al reported that, in 70% of all diagnosed cases of AM, there were images with features of AM before or at the onset of the disease. 16 Although ultrasound has not been recognised as the gold standard of AM diagnosis, the ultrasound features related to AM have been recognised as a consensus, a reliable morphological marker of AM, 17 and can provide supportive evidence for clinical decision-making. Pelvic ultrasound is one routine examination for women of reproductive age for their annual health examination and forms an important part of disease prevention. Health checkups can help screen for potential diseases and profile development trends and the examination results are the main source of information on basic health indicators 18 and can support health recommendations in a timely manner. 19 As there are approximately 500 million health checkups conducted in China, 20 the huge population-based data with ultrasound features of AM enable a more accurate estimation of disease burden and identification of risk factors. We aimed to estimate the prevalence and incidence of age-specific suspected AM in a female Chinese population and to examine the association between biochemical indicators and the presence of AM, using a matched case-control design.

This real-world data analysis was conducted based on the health examination data of the Nanchong Central Hospital, a tertiary hospital in Southwestern China. The health management centre in this hospital conducts approximately 50 000 medical checkups per year. The inclusive criteria of this study were as follows: women with complete clinical and biomechanical data; aged 15–60 years old; undergoing uterus-related examinations with an imaging report (such as abdominal colour Doppler ultrasound, transvaginal ultrasound, or MRI) between October 2017 and December 2020. Women who had a hysterectomy before or during the study period were excluded. This study was a retrospective study that screened the target population directly based on the study’s inclusion exclusion criteria, and informed consent was waived due to the retrospective design. Subsequent dissemination of the results of this study may be conducted through WeChat. We exported the health examination data from the system of the Health Management Center of Nanchong Central Hospital. We used the ID number, age, telephone number and work units to determine whether repeat physical examination data with the same identifiers were of the same person. The detailed data processing flow is shown in online supplemental figure 1 . Finally, we assigned the participants with a unique identification code. Cases of suspected AM were defined as those with reports of ‘adenomyosis considered, suspected, or probable’, verified by two sonographers ( online supplemental figure 2 and table 1 ), according to the diagnostic criteria for AM ( online supplemental table 2 ). Women with endometriosis alone were excluded. Women with no abnormalities in the uterus suggested by the imaging report formed the control group. Multifactorial conditional logistic regression associated with the incidence of AM AM, adenomyosis; CA125, serum cancer antigen 125; CI, confidence interval; DDP, diastolic pressure; HBP, high blood pressure; HR, hazard ratio; SDP, systolic pressure; SE, standard error; TB, total bilirubin. Multifactorial conditional logistic regression associated with the prevalence of AM AM, adenomyosis; BMI, body mass index; BP, blood platelets; CA125, serum cancer antigen 125; CI, confidence interval; DDP, diastolic pressure; HBP, high blood pressure; HDL-C, high-density lipoprotein cholesterol; HR, hazard ratio; SDP, systolic pressure; SE, standard error; TG, triglycerides. Those who did not show any imaging changes of AM on their 2017 examination and had at least two uterine examinations during 2017–2020 were included in the incidence analysis. If any imaging examination after 2017 showed AM-related changes, then the participant was considered as a case, which was used to calculate the incidence of AM. For the identification of factors related to the AM incidence, a nested case-control design was conducted in this population. Those who did not show any imaging abnormalities throughout 2017–2020 formed the control group. The case group was matched to the control group according to the same age and an age tolerance error of 0, 1:2. Baseline characteristics from the first examination in this population were included as potential factors related to AM. All women meeting the inclusive criteria were included in the prevalence analysis. Participants with imaging reports of AM at any time during 2017–2020 were defined as cases for the calculation of the prevalence rate. If more than two AM imaging findings were recorded, data from the first-time indicating changes in AM imaging were used. Factors related to the prevalence of AM were identified using a case-control design, with women without any AM-related imaging changes on uterine examination during the study period (2017–2020) as the control group. The analysis was the same as for incidence of AM. Body mass index (BMI) was categorised as >24 kg/m 2 or ≤24 kg/m 2 . 21 Hypertension (HBP) was defined as systolic pressure (SDP) ≥140 mm Hg or diastolic pressure (DDP) ≥90 mm Hg. 22 Total bilirubin (TB) was classified as <18.81 µmol/L and ≥18.81 µmol/L. 23 According to the criteria for lipid levels established by the National Committee for Environmental Protection in 2001, 24 we converted mg/dL to mmol/L and a total cholesterol level of ≥5.18 mmol/L, triglycerides (TG) level of ≥1.695 mmol/L, high-density lipoprotein cholesterol (HDL-C) level of <1.036 mmol/L, and low-density lipoprotein cholesterol level of ≥2.59 mmol/L indicated dyslipidaemia. Carcinoembryonic antigen 125 (CA125) and blood platelets (BP) were used as a continuous variable in the analysis. We used the following AM incidence formula: Incidence (IN)=P/N*100% where P represents the number of new cases of AM in the observation population during the observation period; N, women with at least two uterus-related examinations during the study period. The denominator of incidence density (ID) and number of person years in the observation period was equal to the date of the most recent physical examination suggestive of new-onset AM minus the date of the initial physical examination of that examinee in the cohort. Standardised incidence of AM: incidence*standard population age composition ratio (age composition ratio of the 7th national census data of China) × 100%. The AM prevalence formula was calculated as follows: Prevalence (PR)=P1/N1*100%. where P1 was the number of new and old AM cases during the observation period; N1, average population during the same period; standardised prevalence of AM, prevalence × standard population age composition ratio (age composition ratio of the 7th national census data of China) × 100%. Normally distributed data are described by mean±SD, and non-normally distributed data are described by median±IQR; categorical data were described by n (%). Rank sum test, χ 2 test and Fisher’s exact test were used to analyse continuous data and categorical data, respectively. Missing data were handled by multiple interpolation. Conditional logistic regression was used to analyse the correlation between AM and biomarkers. Given that AM is an estrogen-dependent disease and oestrogen levels decline with age, only people of childbearing age (15–49 years) were included in the analysis of correlated factors for AM. Linear regression and Cochran-Armitage trend tests were used to analyse the trend of CA125 and incidence and prevalence of AM over age, respectively. We further used the onset AM matched data to construct receiver-operating characteristic curves (ROC) of CA125. Areas under the ROC was used to determine the optimal cut-off point of CA125 level for new AM. The optimal cut-off point was defined as the one with the highest specificity among those with sensitivity>0.7. Two-sided p values less than 0.05 were considered statistically different. All statistical analyses were performed using IBM SPSS Statistics V.26 and SAS V.9.4 (SAS Institute).

A total of 79 746 women participated in the physical examination from October 2017 to December 2020, excluding 139 women with a history of prior menopause or hysterectomy; 21 410 women older than 60 years and younger than 15 years; and 16 556 women without a uterus-related imaging examination; and 94 women had uterus CT examination ( online supplemental figure 2 ). A total of 30 629 women were included. The total number of women who had at least two or more uterus-related examinations was 7818 and a total of 877 cases of suspected AM, of which 121 were new cases. The mean age was 45.47 years for potential AM cases and 40.24 years for non-AM women. In our cohort, 0.10% (32/30629) of women had endometriosis alone, and 0.04% (11/30629) had AM combined with endometriosis. Uterine fibroids (including hypoechoic nodules in the uterine pannus) combined with AM accounted for 14.25% of the 877 cases of potential AM. Transvaginal ultrasound diagnosed 78.79% and abdominal ultrasound diagnosed 19.96% of suspected AM cases. Eleven patients with potential AM were diagnosed with both transvaginal examination and abdominal ultrasound. No women were diagnosed with only MRI ( online supplemental table 3 ). For the total population of 16–60 year olds, the overall incidence of AM was 1.55%, with an ID of 93.56/10 000 person years and a standardised incidence of 1.32%, with the highest incidence at age 46–50 (3.48% or 210.79/10 000 person years) and the second highest incidence at age 41–45 (2.27% or 128/10 000 person years). The overall prevalence of potential AM was 2.86%, with a standardised prevalence of 2.35%. The highest prevalence rate (5.90%) was found in women aged 46–50 years old, followed by those aged 41–45 years (5.14%) ( online supplemental table 4 and figure 1 ). Incidence and prevalence of suspected adenomyosis by age group, 1 October 2017–31 December 2020. After deleting all variables as null, there were 16 280 women aged 15–49 years and 93 with new-onset suspected AM. Univariate tests showed statistically significant differences in DDP (p=0.0182), SDP (p=0.0074), HBP (p=0.0029), TB (p=0.0049) and CA125 (p<0.0001). Compared with the control group, the abnormal diastolic blood pressure, systolic blood pressure, TB, HBP ratio and CA125 levels in control group were slightly higher in the suspected uterine AM group ( online supplemental table 5 ). The above statistically different variables were included in the conditional logistic regression analysis, and the results showed that TB≥18.81 µmol/L (HR: 2.129; 95% CI 1.067 to 4.249; p=0.0321), CA125 levels (HR: 1.014; 95% CI 1.002 to 4.731; p=0.0273) were statistically different ( table 1 ). After deleting all variables as null, there were 16 508 women aged 15–49 years and 638 with suspected AM. Univariate tests showed statistically significant differences in BMI (p<0.0001), DDP (p=0.0006), SDP (p=0.0005), HBP (p=0.0003), TG (p=0.0222), HDL-C(p=0.0002), CA125 levels (p<0.0001) and BP (p<0.0001). Compared with the control group, overweight, abnormal DDP, SDP, HBP, TG, HDL-C ratio, CA125 levels and BP levels were higher in the suspected uterine AM group ( online supplemental table 6 ). The above statistically different variables were included in the conditional logistic regression analysis, and the results showed that BMI>24 kg/m 2 (HR: 1.262; 95% CI 1.055 to 1.511; p=0.0109), CA125 levels (HR: 1.007; 95% CI 1.006 to 1.009; p<0.0001), BP levels (HR: 1.002; 95% CI 1 to 1.003; p=0.0141) were statistically different ( table 2 ). In the new suspected AM case-control group, before the age of 35 years, CA125 levels were lower in the new-onset suspected AM group than in the control group, whereas after the age of 35 years, CA125 levels were higher in the new-onset suspected AM group than in the control group. There was no trend change in CA125 levels with age in either the new-onset suspected AM group or the control group (P trend=0.1275, P trend=0.1147, respectively). In the 15–49 years old group, the incidence and prevalence of suspected AM increased with age (P trend=0.0048) ( online supplemental figure 3 ). CA125 is often influenced by age. To further explore whether the effect of age on CA125 influences the exploration of factors associated with suspected AM, we stratified the analysis by 40 years old, and the results remained consistent between the unstratified and stratified groups. The details were shown in the online supplemental tables 7 and 8 . The ROC of CA125 for the new suspected AM showed that the optimal cut-off value was 10.714 U/mL, with a sensitivity of 77.42%, specificity of 53.76% and AUC of 0.7841 (95% CI 0.7276 to 0.8407) ( figure 2 ). Receiver-operating characteristic curve (ROC) of cancer antigen 125 for the new adenomyosis. AUC, areas under the receiver-operating characteristic curves.

Using the health examination data, we estimated the incidence and prevalence of potential AM in Chinese women based on trans-vaginal ultrasound, abdominal B-ultrasound, or MRI, the most used invasive approaches in gynaecological practice. Furthermore, we identified correlated factors for AM, supporting the preventive plan for disease progression. The association between CA125 and new potential AM in the population level provided a new threshold to correlate the biomarker and disease, which will help provide a more accurate judgement of potential AM. Our study evaluated a standardised incidence of 1.32% and a standardised prevalence of 2.35% for suspected AM. The standardised incidence we evaluated was slightly higher than in the US study. 16 The standardised prevalence of AM we assessed was much lower than that of Naftalin et al , 2 but higher than that of Loughlin et al . 25 Despite all these studies used imaging for AM screening, such variability is unsurprising, likely attributed to the differing study populations chosen: hospitalised populations, 16 25 and women with symptoms in gynaecological clinics. 2 The physical examination population, on the other hand, may contain asymptomatic women in addition to these aforementioned populations. Consequently, AM prevalence as determined by physical examination data might align more closely with its true prevalence. As one of the few studies in China to use medical health examination data, our study provides new support for a more accurate assessment of the disease burden of AM. We found CA125 is a potential marker for potential AM, consistent with previous studies. 26 Currently, the true pathogenesis of AM is unknown; one of the currently accepted theories is ‘Tissue injury and repair’. 27 A clinical manifestation of AM is abnormal uterine bleeding, and bleeding is a major marker of tissue damage. 28 The process of tissue damage and repair causes a series of inflammatory responses and immune regulation. If there is inflammation in the abdomen, whether malignant or benign, CA125 levels will be elevated. 29 CA125 is commonly used in the investigation of ovarian cancer in symptomatic women visiting primary care facilities worldwide. 30 Extensive research has been conducted on CA125 in screening studies, particularly targeting women under 50 in secondary care with pelvic masses, with testing being performed in 39% of this demographic. 31 Although it appears unlikely that CA125 alone would be a useful biomarker for detecting other benign and malignant adnexal masses, 32 high CA125 levels in older patients necessitate further examination to identify the cause and determine the need for further testing. Our findings revealed a correlation between a high level of TB and AM. Previous studies on the role of oestrogen in the development of conditions characterised by high TB, such as chronic hepatitis and liver cancer, have yielded inconsistent results. 33–38 There has been no direct evidence confirming a relationship between TB and oestrogen. Our results, however, suggest a potential correlation between these two biomarkers. Future studies could hypothesise about the role of TB in the development of AM. This implies that TB might serve as an early indicator of AM. We also found that BMI was positively associated with the prevalence of AM, consistent with previous studies. 39 40 The contribution of BMI to the causal pathway cannot be inferred when BMI is measured at the time of AM diagnosis or after diagnosis, that is, BMI may actually coincide with AM, or it may be a consequence of AM. Meanwhile, a previous study showed that oestrogen affects the distribution of body fat, but BMI does not represent the distribution of body fat. 41 Therefore, further studies on the relationship between BMI and AM in this study are needed to demonstrate. A previous study showed that tissue factor (TF) immunoreactivity of the endometrium was significantly increased in women with AM who had heavy menstrual bleeding and dysmenorrhoea. 42 Elevated levels of procoagulant factors such as TF in patients with endometriosis. 43 One of the physiological roles of platelets is to stop bleeding and clotting; similarities exist between AM and endometriosis in terms of disease definition and pathophysiology 44 ; therefore, there may be a relationship between platelets and prevalence of AM, but further studies are needed to prove it. In this study, we found that AM combined with endometriosis was less in 1.25% (11/877) of patients with AM undergoing uterine examination. Additionally, AM often co-occurs with uterine fibroids. We found AM combined with fibroids or hypoechoic nodules in 14.25% (125/877), which is lower than the study of Naftalin et al . 2 A study investigating endometrial hyperplasia reported a strong positive correlation between endometrial hyperplasia and AM. 9 However, this study did not explore the comorbidities of AM and, in the future, we will conduct a multicentre, prospective study to include more samples to further explore the factors and comorbidities associated with AM and infer a causal relationship between them. Furthermore, we identified a new CA125 threshold associated with uterine AM that is lower than previous findings. 45 Although CA125 alone has not been determined to be of high value in the diagnosis of AM, Liao et al found that three-dimensional transvaginal ultrasound combined with CA125 has important diagnostic value in the diagnosis of benign and malignant endometrial lesions. 46 Our study has several limitations. First, this is a retrospective study, and compared with hospitalisation data, physical examination data do not have a medical record to collect more complete information about the symptomatic profile, reproductive history, treatment, comorbidities, and so on, of the patient population with AM. Second, we could not know in which menstrual cycle the CA125 was measured, and we did not know the fluctuation of CA125 during menstruation, which prevented us from controlling for the effect of this condition on AM. Finally, the potential AM filtered out in this study was not confirmed by pathological examination, thus presenting the possibility of overestimation, so our population were limited to women who had uterus-related imaging examinations and could not be generalised to all other women.

The disease burden of AM remains huge. The standardised incidence and standardised prevalence of potential AM were 1.32% and 2.35%, respectively, highest at the age of 46–50 years. In the general population, suspected uterine AM is positively associated with levels of CA125. The optimal cut-off value of CA125 for new suspected AM is 10.714 U/mL, not 35 U/mL. Our data may provide a basis for profiling epidemiological features of AM.