The utility of dynamic MRI in differentiating the hormone-producing ability of pituitary adenomas

Pituitary adenomas are the most common intracranial endocrine tumours and possess various hormonal activities. Dynamic magnetic resonance imaging (MRI) facilitates the detection of pituitary microadenomas [1‐5]; moreover, it is a useful diagnostic procedure for visualising pituitary glands displaced by large pituitary adenomas [1, 6]. Several studies have assessed the relationship between MRI features of pituitary adenomas and hormonal activity [7‐9]; however, they mainly focused on non-contrast MRI findings. Guo et al. [10] performed a quantitative analysis on the contrast patterns of the dynamic MRI images of ACTH-producing pituitary adenomas and found that the tumour prepeak slope (referring to the measure of the tissue perfusion rate or speed reaching peak intensity in the enhancement curve) of ACTH-producing adenomas was significantly lower than that of non-functioning adenomas. In another report by Kanou et al. [11], which analysed the sequential MRI enhancement patterns of pituitary adenomas in 67 patients and the factors contributing to them, it was found that tumours with a very early enhancement pattern were significantly more fibrous than tumours with the other three patterns. However, no significant association between hormonal activity and MRI findings was reported. A recent study reported that dynamic MRI findings can predict the consistency of pituitary adenomas [12]; however, to our knowledge, there has been no statistical comparative analysis of hormonal activity and dynamic MRI findings of pituitary adenomas.

In our institution, all cases of pituitary adenomas, regardless of size, undergo dynamic MRI for preoperative evaluation. We have empirically found that dynamic contrast enhancement patterns of pituitary adenomas range from a greater washout pattern to a prolonged enhancement pattern. Dynamic MRI findings may be useful in predicting hormonal activity in case these enhancement patterns are associated with hormonal activity. This retrospective study aimed to assess the relationship between hormonal activity and dynamic contrast enhancement patterns of pituitary adenomas.

Figure 1 summarises the patient enrolment process. Among the 325 cases of pituitary adenomas, 297 adenomas were pathologically diagnosed at surgery, while 28 PRL-producing pituitary adenomas were diagnosed based on imaging studies and blood hormone levels. PRL-producing adenomas were classified through imaging of the pituitary adenoma and by the presence of serum PRL concentrations exceeding 200 ng/mL. PRL-producing adenomas were also classified through imaging of the pituitary microadenoma and by the presence of serum PRL concentrations between 50 and 200 ng/Ml, because NF macroadenomas may cause mild hyperprolactinemia due to pituitary stalk compression. We excluded 81 pituitary adenomas as follows: 22 adenomas that did not undergo dynamic MRI examinations; 3 GH-producing adenomas and 15 PRL-producing adenomas involving drug treatment before MRI examination; 3 small adenomas (< 5 mm), which impeded assessment through contrast-enhanced imaging; and 38 adenomas with severe cystic degeneration and haemorrhage, which impeded evaluation of contrast images. Among the remaining 244 cases of pituitary adenomas, 55, 16, 6, and 4 produced GH, PRL, ACTH, and TSH, respectively, while the remaining 163 were clinically diagnosed as NF adenomas. The latter lacked pituitary hormone hypersecretion and included multiple histological types, such as silent histology gonadotroph adenomas, silent corticotroph adenomas, and null cell adenomas. Among the 55 GH-producing adenomas, 43 were subclassified pathologically as either densely or sparsely granulated. Of the remaining 12 cases, 3 were subclassified as mixed, while 9 were unclassifiable.

Table 1 summarises the MRI findings. The respective mean values of WR, EER, and DER were 9.4%, 75.2%, and 64.5% for GH-producing adenomas; 6.2%, 117.1%, and 106.2% for PRL-producing adenomas; and 5.4%, 116.7%, and 108.7% for NF adenomas. Dynamic MR images indicated that compared with other adenomas, GH-producing adenomas tended to have a higher WR and a lower EER and DER. A typical example is shown in Fig. 2. Non-GH-producing tumours, including NF adenomas, tended to have a lower WR and a higher EER and DER. A typical example is shown in Fig. 3. The WR was non-significantly higher in GH-producing adenomas than in NF adenomas (P = 0.039) (Fig. 4a). The EER was significantly lower in GH-producing adenomas than in PRL-producing adenomas, NF adenomas, and other adenomas (P < 0.001, P < 0.001, and P = 0.004, respectively) (Fig. 4b). Moreover, DER was significantly lower in GH-producing adenomas than in PRL-producing adenomas, NF adenomas, and other adenomas (P < 0.001, P < 0.001, and P = 0.001, respectively) (Fig. 4c). ROC curves focusing on the differential diagnosis between GH- and non-GH-producing adenomas were analysed using data from WR, EER, and DER. The areas under the curves for the WR, EER, and DER were 0.593, 0.825, and 0.857, respectively (Fig. 5).

On T2-weighted MR images, GH-producing adenomas had lower signal intensities than non-GH-producing adenomas. The T2 signal intensity ratio was significantly lower among GH-producing adenomas than among NF adenomas (P < 0.001); and was also significantly lower in GH-producing adenomas compared to PRL-producing adenomas (P < 0.001) (Fig. 4d). ROC curves between GH- and non-GH-producing adenomas revealed that the area under the curve of T2 signal intensity ratio was 0.814 (Fig. 5d). There were no significant differences in T2 signal intensity ratio between GH-producing adenomas and other adenomas. Densely granulated GH-producing adenomas had a significantly lower T2 signal intensity ratio than sparsely granulated adenomas (P = 0.04). Contrastingly, there was no significant difference in WR, EER, and DER between densely and sparsely granulated GH-producing adenomas.

Our findings indicate that dynamic MRI could be used to predict GH-producing adenomas in pituitary adenomas. Compared with non-GH-producing adenomas, GH-producing adenomas showed a weaker enhancement pattern on the early and delayed phases. ROC analysis for differentiating GH-producing from non-GH-producing adenomas showed high AUC values of 0.825 and 0.857 for EER and DER, respectively, with a specifically high diagnostic performance for DER. Consistent with the findings by Hagiwara [9], we found that GH-producing adenomas had a lower T2 signal intensity ratio than PRL-producing adenomas and NF adenomas. However, there was no significant difference in the T2 signal intensity ratio for other adenomas. These findings suggest that dynamic MRI could allow differentiation of GH adenomas from other adenomas, which was not adequately allowed by previous T2WI findings. Furthermore, the AUC value of the T2 signal intensity ratio was 0.814, indicating that the AUC value (0.857) for DER was higher than that of the T2 signal intensity ratio.

Surgery for GH-producing adenomas requires more complete tumour removal than surgery for non-functioning adenomas; moreover, preoperative diagnosis is important. In most cases, the diagnosis of acromegaly can be made by observing clinical symptoms and the presence of GH or Insulin-like growth factor-1 (IGF-1) in the blood. However, in some cases, clinical symptoms are insufficient, and diagnosis is difficult. On the contrary, some cases of low GH acromegaly show low GH levels in the blood but present clinical symptoms, whereas IGF-1 levels in the blood may be low in those with a low nutritional status, liver disorders, and poorly controlled diabetes mellitus. Our MRI findings may be useful for clinical diagnosis in such cases. Moreover, it may be useful to differentiate GH-producing adenomas by imaging diagnosis in cases of pituitary apoplexy or severe visual dysfunction that require urgent or quasi-urgent surgery, because the IGF-1 blood test is often outsourced and generally has a long turn-around time.

In this study, GH-producing adenomas were smaller than NF adenomas. Given that the tumour size might be associated with dynamic MRI findings, we analysed the correlation between tumour size and dynamic MRI features (WR, EER, and DER). None of the variables were found to be closely correlated with tumour size. This suggests that the pattern differences across the adenomas were not associated with tumour size; however, they could be attributed to differences in other histological structural features.

Vidal et al. [13] reported that GH-producing adenomas had a significantly lower microvascular density than PRL-producing and NF adenomas, which is consistent with the lower EER and DER for GH-producing adenomas than for other adenomas. Di et al. [14] reported that functional adenomas showed a non-significantly lower microvascular surface area than non-functional adenomas. Similarly, this is consistent with our finding that GH-producing adenomas had lower EER and DER values than the NF adenomas. Takano et al. [15] reported that GH-producing macroadenomas tended to have significantly and non-significantly smaller vessel diameters and circumferences than PRL-producing macroadenomas and NF macroadenomas, respectively. This is also consistent with the lower EER and DER in GH-producing adenomas than in PRL-producing adenomas and NF adenomas.

Densely granulated GH-producing adenomas are often effective with somatostatin analogues and have a lower signal on T2WI [9], which is consistent with the lower T2 signal intensity ratio of densely granulated adenomas than in sparsely granulated adenomas. Wei et al. [16] reported that pituitary adenomas with a lower T2WI signal ratio of the tumour and pons had a higher collagen content. Moreover, differences in dynamic MRI contrast patterns were expected to result from histological structural differences. However, we found no difference in the WR, EER, and DER between densely and sparsely granulated types.

Dynamic MRI is useful for detecting pituitary microadenomas, which are difficult to identify on non-contrast MRI, since the normal pituitary gland is very strongly contrast-enhanced, while adenomas show a relatively low intensity. However, we found that GH-producing adenomas had lower contrast enhancement than PRL-, ACTH-, and TSH-producing adenomas. This suggests that the latter adenomas could yield inferior contrast between the adenoma and normal pituitary gland compared with GH-producing adenomas when examining for microadenomas on dynamic MRI. Moreover, ROC analysis for differentiating between GH-producing and non-GH-producing adenomas showed high diagnostic performance, especially for DER. This suggests that it may be better to focus on the delayed phase, rather than the early phase when detecting and diagnosing GH-producing adenomas on dynamic MRI.

Dynamic MRI could be used to predict GH-producing adenomas in pituitary adenomas. Compared with NF adenomas, PRL-producing adenomas, and other adenomas, GH-producing adenomas had a weaker enhancement pattern on the early and delayed phases. In addition, compared with GH-producing adenomas, PRL-, ACTH-, and TSH-producing adenomas could yield inferior contrast between the adenoma and normal pituitary gland when searching for microadenomas on dynamic MRI. It may be better to focus on the delayed phase, rather than the early phase for detecting and diagnosing GH-producing adenomas on dynamic MRI.