DISCUSSION:
Main Findings:
We assessed clinical and MRI parameters for their potential to predict histopathological adenomyosis diagnosis prior to hysterectomy. The resultant multivariate prediction model discriminates well between patients with and without adenomyosis (AUC 0.776). Five clinical characteristics: age at MRI, BMI, history of curettage, dysmenorrhoea, and AUB, and four primary MRI parameters: mean JZ thickness, JZ Diff ≥5 mm, JZ/MYO >.40, and the presence of HSI foci are included.
Strengths and Limitations:
This study has several strengths and limitations that merit consideration. A strength of our study is that two researchers independently reviewed all pelvic MRIs blinded to the histopathology outcome. Furthermore, the proposed model was built on data of 296 patients and data driven variable selection was avoided, along with corrections for potential overfitting. Additionally, the combination of both clinical and MRI parameters makes this model easily implementable into daily clinical practise.
The present study used broad inclusion criteria, which could be interpreted as both a strength and limitation. On the one hand, inclusion of patients with comorbidities like uterine fibroids might have prevented an overestimation of diagnostic performance of the individual potential predictors. Alternatively, severe distortion of the uterus due to fibroids or endometriosis can limit the ability for objective assessment of all MRI parameters.
One limitation of the current study is that it was not possible to correct for the influence of the menstrual cycle on MRI parameters. Although it is known that JZ thickness changes during the menstrual cycle 25, cycle phase at time of MRI was not reported for most of our patients. Furthermore, the choice for histopathology after hysterectomy as a reference standard introduces an element of selection bias. Potentially, our group consisted of women with more severe adenomyosis and thus may have affected the general phenotype. The present study did not conduct a central review of pathology however, and (histological) adenomyosis severity was generally not reported in pathology reports. Therefore this remains hypothetical.
Interpretation of findings:
To the best of our knowledge, no comparable models for histopathological adenomyosis diagnosis based on MRI exist. Previous studies have investigated prediction of adenomyosis diagnosis based on ultrasound, with comparable accuracy 9,11,26. However, it is known that ultrasound diagnosis is highly operator dependent, with varying inter- and intra-observer variability 8,27,28. An MRI prediction model such as developed in our study thus has clinical value especially in cases where adenomyosis co-exists with other pathology (as was the case in the majority of our included patients), or is mild, or atypical.
The parameters ultimately included in this model are unsurprising when considering reported adenomyosis clinical presentation and aetiology. Dysmenorrhoea and AUB are the most frequently reported symptoms of adenomyosis 4,29 and were thus logical (and statistically significant) additions to the model. Age at MRI was further included in the model due to the known physiological increase in JZ thickness with age 25,30,31. BMI was also manually entered into the model as, despite univariate analysis showing no significant association, increased body weight and obesity have been reported as strong risk factors for adenomyosis 32.
History of curettage (after miscarriage) established itself to be an important predictor and was thus included in our model. It is debatable as to if curettage is a cause or a consequence of adenomyosis, as adenomyosis is often associated with risk of miscarriage5. Conversely, curettage as a risk factor for the development of adenomyosis could potentially be explained by iatrogenic trauma leading to the mechanical transport of endometrial cells into the myometrium 33,34.
None of the primary MRI parameters alone were sufficient to diagnose adenomyosis conclusively, which is in line with the literature (11) . The presence of HSI foci emerged as the strongest predictor of the assessed MRI parameters (p <.001). Bazot et al. indeed described these foci as the only direct diagnostic criterion and almost pathognomonic for adenomyosis on MRI, although they are detected in only about half the cases (11).
In clinical practice, our model could be used to calculate the risk of adenomyosis in individual patients. For example, in a 31-year-old woman with a BMI of 19 kg/m2, without history of curettage, with complaints of both dysmenorrhoea and AUB, and an MRI with mean JZ thickness of 8.3 millimetres, a JZ Diff <5 mm, a JZ/MYO >.40, but HSI Foci (Figure 1A), the probability of adenomyosis is 14.9%. In a 35-year-old woman with a BMI of 24 kg/m2, without history of curettage, with complaints of both dysmenorrhoea and AUB, and an MRI with a mean JZ thickness of 24.6 millimetres, a JZ Diff ≥5 mm, a JZ/MYO >.40 and HSI Foci (Figure 1B), this probability increases to 90.3%.