2 Designing and Reporting Studies on Intraoperative Tissue Imaging in the Brain

Surgery for gliomas has changed considerably during the past two decades. The value of safe maximal resection in all types of gliomas is well established1,2,3 as reflected by neuro-oncological guidelines.^4,5 Methods of intraoperative mapping and monitoring are more and more being employed to ensure maximal safety.⁶ However, regarding the process of resection, it is commonly acknowledged that the surgeon’s visual perception of tumor-infiltrated brain, affording at best subtle variations of color and texture compared to normal brain, and the haptic impression of infiltrated brain are limited in their utility for identifying tumor⁷ even when using the surgical microscope. Due to inevitable neurologic morbidity, operating with a safety margin within the brain is seldom an option. Exact identification of tissue dignity is of utmost importance for improving oncological outcome, whereas neurological function has to remain intact.

This realization has spawned methods for intraoperative identification of residual tumor, such as the intraoperative MRI,^8,9,10 ultrasound,^11,12,13 or neuronavigation.^14,15,16 All of these techniques have inherent limitations, such as the price for MRI or brain shift for navigation, and none of these methods are truly real time. All require interrupting surgery to collect more or less accurate information on where infiltrated brain might be located. This information then needs to be transposed into the surgical cavity before the surgeon can resume surgery. Highly precise methods such as intraoperative histology from frozen sections, or newer methods such as confocal microscopy,^17,18 autofluorescence spectroscopy,^19,20 or Raman spectroscopy^21,22,23 have a high resolution and exact spatial allocation. However, only small regions of the cavity can be assessed at one time, making these techniques useful for “near-time” intraoperative biopsies but limiting their usefulness for real-time resection in a macroscopic environment.

Due to these limitations, it is clear that the new and expanding field of intraoperative optical tissue diagnosis is attractive to neurosurgeons. Optimally, techniques of this type would allow direct visualization of tissue under the surgical microscope.

The advantages offered by optical tissue imaging are many ( Table 2.1). Information on tissue dignity can be obtained in real time, and surgical decisions can be made immediately, without having to interrupt surgery, since the surgeon is seeing the signal (e.g., fluorescence) through the microscope in the tissue while he or she is operating. Tissue manipulation and tumor resection are possible while utilizing the additional optical information at the accustomed magnification. Brain shift, a confounder in neuronavigation, is not a worry with visual methods, and the optical technology is integrated into the operating microscope and does not require additional equipment in already overfilled operating rooms. Furthermore, using the adapted surgical microscope, the surgeon is operating with his or her familiar magnification on the visual aid he or she feels comfortable with after years of training.

There are several potential drawbacks however. Restricting the optical signal coming from the tissue might also result in loss of visual information. This is particularly true for fluorescence. Since only selected wavelengths of excitation light are employed, the background is generally not nearly as strongly illuminated as accustomed. While this problem can be partially mitigated by toggling between conventional and fluorescence illumination, it still hinders unimpaired surgery. Furthermore, methods of intraoperative tissue diagnosis will typically offer only surface information in a two-dimensional fashion, and the tissue signal might be obscured by blood or hemostatic agents, such as Surgicel. Strong light might lead to bleaching of signal in case of dyes that are used,^24,25 and surgical disruption of the blood–brain barrier might result in extravasation of agents carried by blood.^26,27

Table 2.1 Potential advantages and disadvantages of intraoperative optical tissue imaging

In its present form, intraoperative tissue diagnosis was spawned by the introduction of porphyrins induced in tumor tissue by 5-aminolevulinic acid25,28 (5-ALA), which can be visualized with appropriate excitation light and filter systems incorporated into the microscope. 5-ALA remains the only approved agent for the purpose of intraoperative tissue imaging. After approval of 5-ALA in Europe by the European Medicines Agency in 2007 and by the Food and Drug Administration (FDA) in 2017,²⁹ this field has gained further momentum. ALA, as approved for brain surgery, is thus the only regulatory precedent at present. However, many new methods are being explored using different fluorochromes, which are either non- targeted or targeted to malignant tumor tissue.³⁰

2.2 Regulatory Perspective

From the FDA perspective,³¹ approval of an agent is principally linked to two aspects: safety and benefit to the patient. Endpoints that simply correlate a tissue signal, for example, fluorescence, with the location of a known tumor may not be considered sufficient for approval. Furthermore, for approval, cost-effectiveness is primarily not an issue, as the responsibilities of FDA when evaluating technologies and agents do not initially focus on costs. This is an issue for the Centers for Medicare & Medicaid Services (CMS) at a later stage of the introduction of an agent.

Regulatory issues also pertain to the hardware side of technologies for intraoperative imaging. Approved devices might be tested in conjunction with new agents, or new devices might have to be tested with known agents, with the intent of improving imaging from the hardware side using the known agent. This chapter does not cover possible requirements for marketing application of device clearance by the FDA nor how such devices might be tested in detail or compared.

An agent for which approval might be considered will have to demonstrate a benefit to patients. Benefit could be based on the additional identification of diseased tissue after standard surgical resection. This, however, assumes that a more complete removal of malignant tissue is directly correlated with survival or other clinical benefits, such as a reduced need for reoperation. A correlation between histologically clear margins and survival is already well established for many types of cancer. The same may be true for debulking surgery. For example, if prior evidence indicates that debulking correlates to better outcomes, it may be sufficient to show that the optical technique improves the surgical safety and effectiveness of debulking procedures.³¹ In the context of gliomas, regulatory bodies have principally questioned these points. This aspect was a central point of discussion with the FDA in the process of gaining approval for 5-ALA, the reason being the paucity of randomized studies in neurosurgery addressing these issues, despite the conviction of the neuro-oncological community.

One major issue during the conduct of imaging studies is the minimization of bias.³¹ Randomization might be a solution; however, blinding the operating surgeon is obviously not an option. Intraoperative controlled studies might be another option, that is, using one method (conventional illumination), and then the new method in a controlled way. Again, such an approach would also have to be well designed to diminish biases, which are inherent to any optical method.

Table 2.2 Development steps for brain intraoperative imaging agents or technologies

The first obvious step in assessment of a method—after establishing toxicological safety and pharmacokinetics—would be to verify the diagnostic performance of that method. Does the method truly show tumor and to what extent it is capable of showing tumor-infiltrated brain ( Table 2.2)? Such data might first be generated by animal experiments and subsequently in patients.

Later steps might be to determine how the tissue signal relates to the tumor as depicted on the MRI or to establish methods of preoperative tumor imaging. In this context, it must be remembered that gliomas are diffusely and extensively infiltrating lesions. Expecting a method to find every single tumor cell would not be feasible in the context of glioma surgery. Typically, the extent of visualizing low-grade glioma would be the MRI fluid-attenuated inversion recovery (FLAIR) anomaly, and in high-grade gliomas the gadolinium-enhancing portion of the tumor.

Finally, it would be convincing to demonstrate that the use of these agents will translate into improved clinical outcomes, such as enhanced, safe resections or prolonged progression-free and overall survival, the ultimate goal of surgical oncology. On a bynote, the FDA does not specifically require survival as an endpoint for demonstrating the effects of an intraoperative imaging technique.³¹

Importantly, given the plethora of agents and procedures for intraoperative tissue imaging that are being developed, the fundamental question will arise on how these methods perform when they are directly compared. Methods might compete based on price and utility, but ultimately patient safety and efficacy will prevail as the most important factors characterizing an agent and and lead to its preference over others.

For this reason, it is clear that the community requires standards for assessing intraoperative imaging methods. Little precedent is available so far, and to our knowledge the FDA offers no guidance regarding this new field.

The last part of this chapter will discuss aspects related to the assessment of the diagnostic performance. In particular, this chapter will give an overview of perceptions and possible misconceptions regarding the assessment and reporting of intraoperative methods of imaging. Since this is an emerging field without detailed regulatory precedent, the academic community should help in defining possible approaches to these issues as soon as possible.

2.3.2 Possible Variables and Pitfalls Related to Assessments of Diagnostic Accuracy in Intraoperative Imaging

• Thresholds and their influence on diagnostic accuracy. Some methods will also highlight normal tissue to a certain degree. This would be expected for methods that rely on intrinsic signals, for example, autofluorescence19,20 and would also be expected for technologies that use agents for highlighting tumors that circulate in the vessels after being administered intravascularly, for example, fluorescein.^26,27,32,33 These compounds will highlight all brain tissue to a certain extent, despite having higher concentrations in tumor tissue. Especially low signal-to-noise ratio methods will require some form of threshold definition, above which a sample is considered positive. This threshold will have to be clearly described. Increasing the threshold will lead to fewer false negatives, while lowering the threshold will lead to more false positives. Thus, defining a threshold will directly affect diagnostic accuracy. Optimally, receiver operating curves³⁴ should be employed to relate all thresholds to their respective diagnostic accuracies. Again, it is important that thresholding will only play a role with methods that rely on imaging processing. With visual methods, defining thresholds will not be an option. Human visual cognition is too flexible and adaptable to allow objective thresholding.

• Changes in intraoperative signal over time. Researchers should be mindful that during a neurosurgical procedure lasting several hours, the signals induced in tissue will change over time. This is the case for ALA-induced porphyrins with a maximum at about 8 hours.³⁵ With intraoperative dyes such as fluorescein, the situation is somewhat more complex.^27,36,37 While fluorescein has no specific affinity to tumor cells, concentrations are at first high in perfused tissue and are later extravasated in tumor, leading to pseudoselectivity. However, there is some concern about unspecific propagation with edema.³⁸ Reporting timing here is essential.³⁹ With “intelligent” fluorophores, they are selectively retained in tumors, for example, alkylphosphocholine (APC) analogs,⁴⁰ fluorochromes conjugated with chlorotoxin,^41,42 or fluorochromes targeting integrin receptors.⁴³ These will first be in the plasma and will enter the tumor via a breached blood–brain barrier, but will selectively be retained due to the specific interaction or binding of the fluorochrome to the tumor cells. Unspecifically, extravasated surplus dye molecules will be washed out of the brain interstitial space and dye will be cleared from plasma with time. The results on diagnostic accuracy will thus have to be reported in conjunction with the time after application of a compound.

• Truth standards. In order to determine the diagnostic accuracy of an agent for correctly identifying the dignity of tissue, generally acknowledged comparators need to be defined on which to base this assessment. The most obvious comparator is histology. However, especially in diffusely growing tumors it may not always be possible for neuropathologists to correctly identify single tumor cells. Molecular markers, such as isocitrate dehydrogenase 1 (IDH1) mutations, Ki-67/MIB-1, or p53 might increase the sensitivity of actually finding tumor cells as opposed to simple hematoxylin and eosin (H&E) staining. Since the sensitivity of detecting tumor cells will directly influence the reported diagnostic performance of a method, the methods used need to be reported exactly.
Established methods might serve as comparators or truth standards. For the randomized 5-ALA study, the comparators were resection rates using conventional microsurgery with white light. Future studies might require resection rates using 5-ALA fluorescence-guided resection to be a comparator. Another possible truth standard might be MRI, since our assessments of the completeness of resection and disease status are based on this modality. Using MR as a truth standard would imply an exact correlation between intraoperative signal and MRI, for example, by neuronavigation. Resolution issues and brain shift may be distinct confounders that need to be accounted for. Table 2.3 compiles several truth standards, the type of study they might be used in, and possible confounders related to such truth standards.

Table 2.3 Possible comparators or truth standards

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

Intraoperative Fluorescence Guidance in Meningiomas 5-Aminolevulinic Acid and Recurrent High-Grade Gliomas Current Fluorescence-Guided Neurosurgery and Moving Forward Confocal Endomicroscopy Indocyanine Green Videoangiography and Arteriovenous Malformations Cancer-Targeted Alkylphosphocholine Analogs for Intraoperative Visualization

Stay updated, free articles. Join our Telegram channel

Join