Is Contrast Medium Really Needed for Follow-up MRI of Untreated Intracranial Meningiomas?

Individual Measures and Tumor Growth

Our study demonstrates that the absolute agreement among parameters obtained from the T1 3D-Gd and T2WI sequences with respect to tumor volume, maximum 3D diameter, maximum 2D diameters in 3 planes, and major axis length was near-perfect, despite T2WI sequences being 2D with 4-mm section thickness and a 0.4-mm gap between slices, demonstrating the value of such noncontrast sequences for the follow-up of untreated meningiomas. This feature proved to be the case even across a variety of 9 scanner models from 2 different manufacturers in our institution and also a third scanner manufacturer from outside institutions, thus reinforcing the robustness of the agreement we found between T1 3D-Gd and T2WI segmentations.

An important point is the subgroup analysis we performed to investigate whether the agreement between the 2 sequences remained acceptable if only certain types of meningiomas were considered. The agreement between T1 3D Gd and T2 remained adequate for meningiomas in different locations, smaller-versus-larger meningiomas, and also symptomatic-versus-asymptomatic lesions for all the different single-time-point measurements, including the most important ones such as tumor volume, maximum 3D diameter, and maximum diameter in all 3 planes. The subgroup analysis also found that symptomatic meningiomas tended to be larger than asymptomatic lesions, as expected. These high ICC values provide, in our opinion, further reassurance that our method was consistent and that noncontrast MR imaging could potentially suffice for the follow-up of a wide variety of meningiomas, including symptomatic lesions. We understand, however, that the case could be made for the use of Gd contrast media for a small proportion of aggressive and infiltrative meningiomas in very specific locations, which could hinder an appropriate imaging work-up, at least in the first scan.

The growth rate of intracranial meningiomas is one of the most important factors influencing the decision for surgical treatment, with immediate intervention at diagnosis having the potential for unnecessary treatment.³⁰ Furthermore, because most asymptomatic meningiomas do not exhibit growth, the best management option may be conservative treatment with imaging follow-up, thus avoiding surgery-related morbidity.²⁵ Serial imaging allowed us to evaluate tumor growth. A conscious decision was made to include only meningiomas in which the follow-up was at least 2 years, to minimize errors as much as possible. Absolute tumor growth (evaluated as an annual increase in tumor volume, tumor surface area, maximum 3D diameter, and major axis length) obtained from the T1 3D-Gd and T2WI segmentations showed excellent agreement between the 2 sequences with near-perfect ICCs. The agreement for relative annual tumor growth expressed as %/year was slightly lower, probably reflecting that more computation was needed to obtain relative annual growth, but it was still very high with P < .001 in all cases. Because MR imaging follow-up of intracranial meningiomas is primarily performed to assess tumor growth, our results are both encouraging and reassuring, indicating that Gd-based contrast might not add enough value and could potentially be omitted.

An important point is that contrast medium is usually administered in follow-up MR imaging of intracranial meningiomas because it is thought to allow improved delineation of invasion of the dural venous sinuses by a lesion, such as the sagittal, straight, or cavernous sinuses. Nevertheless, we found that visualization of vascular invasion by meningiomas is often hindered on T1 Gd-weighted sequences, in which enhancement of the meningioma is sometimes difficult to distinguish from opacification of the venous sinus itself, and that vascular invasion is often better depicted on T2WI as an absence of flow void and by direct visualization of the edges of the meningioma itself. This is the case with gradient-echo T1 3D-Gd sequences, which are routinely performed as part of the meningioma follow-up imaging protocol at our institution. While it is true that black-blood Cube (GE Healthcare) or sampling perfection with application-optimized contrasts by using different flip angle evolution (SPACE; Siemens) sequences would mitigate this limitation, they would also depict other anatomic structures in less detail, especially near the skull base, including the meningioma itself and brain tissue. Image-acquisition time would also be prolonged with this type of sequence. Another important point in this respect is that while it is true that the absence of flow voids on T2WI does not always reflect sinus invasion by the meningioma, dural sinus invasion could be evaluated with other noncontrast sequences. Because meningiomas are usually hyperintense on diffusion-weighted imaging due to their high cellularity, this sequence could be used to distinguish slow flow from sinus invasion.

Because one of the advantages of not using contrast media is to reduce imaging time, it seems somewhat counterproductive to include sequences that would increase the image-acquisition time for little gain in quality. At our institution, the actual gain in imaging time if no contrast was used amounts to the time it takes to perform 2D-T1 spin-echo Gd and T1 3D-Gd sequences, which is approximately 8 minutes depending on the scanner. The time savings would be even more striking if we obviated the need for contrast MRV, a sequence usually included in the imaging protocol of meningiomas that invade or compress the dural venous sinuses. However, 3D-T2 sequences can sometimes depict the venous sinuses and structures near the skull base in less detail due to artifacts. In view of these limitations, a sensible compromise could be to keep 2D-T2WI in the imaging protocol and replace the 2D-FLAIR with a 3D-FLAIR sequence.

As mentioned before, health care cost is also an important factor in modern medical practice, not only to patients in private or semi-private health care systems but also to societies. In our center, the direct cost-saving of performing a noncontrast as opposed to a contrast brain MR imaging would be approximately $500. Indirectly, savings would be even more significant if the reduced image-acquisition time is taken into account. Last and despite the relative safety of Gd contrast media, the short-term and long-term consequences of the administration of contrast media could be avoided and patient safety could be improved.

Literature

The excellent agreement between absolute annual tumor growth obtained from the T1 3D-Gd and T2WI segmentations strongly suggests that noncontrast MR imaging is adequate for the follow-up of most intracranial meningiomas. Our results are in agreement with those by He et al,³¹ in 2020, who assessed the suitability of T2WI sequences for long-term follow-up of asymptomatic convexity meningiomas. However, our study differs from this one in several points, specifically a larger number of patients (123 versus 18), also including skull base meningiomas as well as both symptomatic and asymptomatic meningiomas, and our study tumor size being assessed by a semi-automated method, thus minimizing interoperator error.

The behavior of skull base meningiomas is different from that of non-skull base meningiomas, and skull base meningiomas have shorter retreatment-free survival.²⁸ Notwithstanding some difficulties we encountered with the segmentation of skull base meningiomas in certain locations (in the cavernous sinus and near the orbits), we were reasonably confident that our method had adequate accuracy for all intracranial locations, and we also included skull base meningiomas in our study, which are known to behave differently and exhibit faster growth rates that non-skull base meningiomas as observed by Hashimoto et al,³² in 2012. The fact that the agreement between the two different sequences remained acceptable for skull base meningiomas provided some validation to the decision of including these meningiomas in our study.

Strengths and Limitations

We believe that one of the strengths of our study is that an accurate semi-automated segmentation method was considered from the early planning stages of this study to minimize subjectivity and operator-related variability as much as possible. Furthermore, we included parameters that are not generally used in routine clinical practice such as the least axis length and minor axis length of the tumor, to further confirm the accuracy of the segmentation and the measurements. Another strong point, in our opinion, is that we did not exclude skull base or symptomatic meningiomas, making our results more generalizable and holding higher external validity then if we had focused only on convexity meningiomas.

The main limitations of this study are its retrospective rather than prospective nature and also the potential bias that was introduced by the segmentation process, which was not blinded between the T1 3D-Gd and T2-weighted images, as each meningioma was segmented by the same operator. The decision to perform the segmentation in this manner was made at the beginning of the study in order to allow the considerable workload to be shared and to expedite the segmentation process. Another limitation is that although some MR imaging studies from external institutions were included, it remains a single-center study. Other limitations relate to the technical aspects of the semi-automated segmentation; we found that delineation of skull base meningiomas in certain locations posed considerable difficulties due to adjacent anatomic structures with similar signal intensities. Although care was taken to perform the segmentation of these meningiomas as thoroughly as possible, we acknowledge that some inaccuracies may have occurred while performing the segmentation of meningiomas of the cavernous sinus, both on T1 3D-Gd and T2WI, and those near the orbits on T2WI. A case of a meningioma for which the segmentation was particularly challenging is illustrated in Fig 5. T2WI-based segmentation tended to overestimate tumor growth. This feature was probably due to the effects of partial volume averaging, which are to be expected when a 2D sequence is used to generate 3D parameters, particularly on lesions that are irregularly shaped. We believe this overestimation could, however, be easily overcome if a 3D volumetric T2-weighted sequence was used, which could be a topic of future research.