Intraoperative near-infrared imaging with receptor-specific versus passive delivery of fluorescent agents in pituitary adenomas

Study Design

Patients were enrolled in one of 2 sequential, separate clinical trials (June 2015–January 2017), one evaluating SWIG in intracranial tumors and the other a pilot study evaluating OTL38 administration in pituitary adenomas. Both studies are registered with clinical trials.gov; the registration number for the SWIG study is NCT03262636, and that for the OTL38 study is NCT02629549. Inclusion criteria were diagnosis of a pituitary adenoma and age > 18 years. Pregnancy and history of allergy to folate (OTL38 study) were the main exclusion criteria. Between October 2015 and January 2017, all patients were enrolled in the OTL38 trial; otherwise, patients were enrolled in the SWIG trial. The 2 prospective cohort studies were each approved by the University of Pennsylvania institutional review board, and all patients gave informed consent. Patients underwent preoperative MRI of the brain/sella with and without gadolinium contrast, as well as serum measurements of thyroid stimulating hormone (TSH), growth hormone (GH), adrenocorticotropic hormone (ACTH), and prolactin. Those with neither clinical signs of hormone oversecretion nor elevated serum levels were designated as having nonfunctioning adenomas. Patients were informed that enrollment in the study would not substantially change the extent of surgery.

NIR Fluorophore Administration

Patients were brought to the outpatient infusion suite prior to surgery—24 hours before surgery for the SWIG group and 2–3 hours before surgery for the OTL38 group. The patients in the OTL38 group were instructed to discontinue folate supplements or folate-containing multivitamins 48 hours prior to surgery in order to reduce potential interactions with OTL38. Patients were given 25 mg diphenhydramine intravenously as a precautionary measure and then, over the course of 1 hour, they received infusions of either 5 mg/kg of ICG or 0.025 mg/kg of OTL38. The patients’ vital signs were monitored throughout the infusion and for 30 minutes afterward.

NIR Imaging System

In all cases, NIR signal was visualized using 2 different versions of the VisionSense Iridium camera system equipped with a 4-mm endoscope. This system is FDA approved for perfusion imaging in plastic and reconstructive surgery (to assess tissue flap vascularity, for example). Two different excitation laser sources were utilized. For OTL38, the excitation source is a laser tuned to 785 nm and the sensor is filtered for emission in the 800- to 835-nm range. For ICG, the excitation source is a laser tuned to 805 nm, and the sensor is filtered for emission in the same 820- to 860-nm range. The commercially available VisionSense system was recently compared to other NIR visualization systems, and its sensor gain and sensitivity are considered the best among the commercially available systems.^4,5

Surgical Approach and Study Procedure

The endonasal endoscopic transsphenoid approach was performed by otorhinolaryngologists using standard endoscopic technique.^6,32 Upon exposing the sella, the senior neurosurgeon (J.Y.K.L.) used the VisionSense endoscope to record NIR images. The dura was initially kept intact in order to visualize NIR signal through it. Once the dura was opened, NIR fluorescence within the tumor was visualized (Fig. 1). After tumor exposure, specimens were taken from the tumor mass and sent to pathology for histopathologic diagnosis. Surgery then proceeded in the standard-of-care manner without the use of NIR fluorescence.

After the tumor had been resected and the attending neurosurgeon was satisfied that a complete resection had been achieved based on white light visualization, NIR imaging was used to identify areas of residual fluorescence (Fig. 2). Areas were biopsied at the discretion of the senior surgeon, and specimens were coded (yes/no for NIR fluorescence and yes/no for tumor based on the surgeon’s impression) and sent for pathological analysis. The closure was performed jointly by otorhinolaryngology and neurosurgery. Patients were admitted to the intensive care unit following surgery. A postoperative MRI study was performed on the first postoperative day and then after 2, 6, and/or 12 months.

Immunohistochemistry for FRα in OTL38 Patients

Tumor specimens from patients in the OTL38 arm were chemically fixed and embedded in paraffin blocks and sectioned at 5-μm thickness. Immunohistochemistry for FRα was performed using murine monoclonal antibodies against folate receptors (1:20 dilution of NCL-L-FRalpha; Leica Biosystems). In order to quantify FRα expression in each sample, H-scores were calculated by a single neuropathologist (M.N.). The H-score ranges from 0 through 300, with 300 implying strong staining in all cells and 0 implying no staining in any cell.^2,29

NIR Image Analysis for SBR

The NIR imaging data were analyzed after the surgical procedure using the VisionSense software (VSPlayer v1.8.05.01; VisionSense) and ImageJ (https://imagej.nih.gov/ij/; National Institutes of Health) to calculate the signal-to-background ratio (SBR), as described in our previous paper.²¹ In brief, care was taken to visualize samples where the opticocarotid distance was < 50% of the entire screen. It was previously noted that false-positive detection of NIR fluorescence could be minimized by increasing the distance from the endoscope to the tissue. Thus, all initial images were taken with the medial opticocarotid recess spanning less than half of the screen, as our prior work demonstrated this to be the ideal distance.^15,17,21,41 Regions of interest were drawn around the tumor and adjacent normal tissue (sometimes bone or mucosa).

Data Analysis

Statistical comparisons with linear regression, t-tests, and nonparametric tests were performed using Stata 10 (StataCorp, LLC). Two-by-two contingency tables were constructed to calculate the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for the surgeon’s impression, SWIG, and OTL38.

Clinical Data

A total of 39 patients were enrolled between October 2015 and December 2017. Sixteen patients (7 with nonfunctioning, 4 with GH-secreting, 2 with ACTH-secreting 2 with PRL-secreting adenomas, and 1 with a TSH-secreting adenoma) were enrolled in the SWIG trial and 23 patients (14 with nonfunctioning, 6 with ACTH-secreting, 3 with GH-secreting adenomas) were enrolled in the OTL38 trial (Table 1). SWIG was administered at a mean (± SD of 22 ± 2.4 hours before tumor visualization, and OTL38 was administered at a mean of 4.8 ± 1.2 hours before tumor visualization. All patients tolerated ICG or OTL38 infusions without adverse event and no permanent surgical complications were identified.

Folate Receptor Immunohistochemistry

After standard postoperative immunohistochemical staining for pituitary hormones, further immunohistochemistry for FRα was performed on paraffin-embedded tissue sections using the protocol described in Methods.

Non-neoplastic pituitary tissue from 11 specimens demonstrated a mean FRα H-score of 50 ± 40 (median 34). Nine of 23 adenomas in the OTL38 group demonstrated significant FRα overexpression (H-score > 150), and all 9 were nonfunctioning adenomas. The other 14 adenomas (5 nonfunctioning, 6 ACTH-secreting, 3 GH-secreting) demonstrated FRα H-scores < 100.

Intraoperative NIR Fluorescence

The 16 adenomas in the patients who received ICG (SWIG trial) demonstrated NIR fluorescence through the dura that localized to the tumor with a mean SBR of 2.7 ± 0.63; the SBR increased (to a mean value of 4.1 ± 0.72) upon dura opening and direct exposure of the tumor (Fig. 1). NIR wavelength is longer than visible light, and penetration of fluorescent signal through normal dura has been reported by our group.^20–23 When the cases were grouped by preoperative hormonal status, the average SBR with tumor exposure was 4.1 ± 0.63 for the nonfunctioning adenomas and 4.0 ± 0.82 for the functioning adenoma (p = 0.9). A detailed breakdown of NIR SBR grouped By functional status is shown in Table 2. There were no significant differences in SWIG fluorescence signal between the classes of pituitary adenomas. Tumor specimens were further imaged ex vivo, using both gross specimen and microscopic imaging (Fig. 3). Specimens that were fluorescent intraoperatively demonstrated strong fluorescence ex vivo. On microscopy, it was confirmed that areas of tumor demonstrated diffuse, rather than focal, uptake of ICG.

The 23 adenomas in the patients who received OTL38 (folate dye) fluoresced with a mean SBR of 1.6 ± 0.43 through the dura, and the average NIR SBR increased to 2.2 ± 0.88 upon tumor exposure (Table 2). When the cases were grouped by preoperative hormonal status, the 14 nonfunctioning adenomas had a mean SBR of 2.6 ± 0.91, and the 9 functioning adenomas had a mean SBR of 1.7 ± 0.47 (p = 0.013). When the cases were grouped by postoperatively determined FRα expression, the 9 FRα-overexpressing adenomas demonstrated a mean SBR of 3.2 ± 0.53, while the other 14 adenomas with low FRα expression had a mean SBR of 1.6 ± 0.42 (p = 0.0001) (Fig. 4).

NIR Fluorescence Detection of Neoplastic Tissue

A total of 30 specimens were sampled from the 16 patients enrolled in the SWIG trial. Each time, the specimens were labeled by the surgeon as tumor or not tumor based on visual and tactile assessment; the presence of NIR fluorescence in each specimen prior to biopsy was also recorded as yes or no (Table 3). Using postoperative histopathology as the gold-standard diagnosis of each specimen, the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated. Using an NIR SBR cutoff of 2.0 (i.e., NIR signal in tissue is > 2× the background signal), SWIG was 100% sensitive (95% CI 81%–100%) but only 29% specific (95% CI 5.1%–70%) for neoplastic tissue, with a PPV of 82% (95% CI 74%–88%) and an NPV of 100% (95% CI 20%–100%). The preoperative hormonal status of the patient had no effect on intraoperative fluorescence. In nonfunctioning adenomas, SWIG was 100% sensitive but only 33% specific, with a PPV of 83% and NPV of 100%; in functioning adenomas, SWIG was 100% sensitive and only 25% specific, with a PPV of 81% and NPV of 100%.

A total of 48 specimens were sampled from the 23 patients enrolled in the OTL38 trial. Statistical analysis of NIR imaging using OTL38 was performed only on the 28 specimens from the 14 nonfunctioning adenomas, as OTL38 was hypothesized to have utility only in FRα-overexpressing tumors, such as nonfunctioning adenomas. As above, specimens were coded for tissue identity and for presence of NIR fluorescence by the senior surgeon prior to biopsy. Using an NIR SBR cutoff of 2.0, OTL38 was 75% sensitive (95% CI 51%–90%) and 100% specific (95% CI 60%–100%) for neoplastic tissue, with a PPV of 100% (95% CI 75%–100%) and an NPV of 62% (95% CI 43%–77%).

When the analysis was further limited to the 18 specimens from the 9 FRα-overexpressing adenomas, OTL38 was 100% sensitive (95% CI 75%–100%) and 100% specific (95% CI 31%–100%) for neoplastic tissue, with a PPV of 100% (95% CI 75%–100%) and an NPV of 100% (95% CI 31%–100%).

Overall, 78 specimens from 39 patients were sampled. The senior surgeon’s determination of tissue identity based on white light, visual, and tactile impression was 88% sensitive (95% CI 75%–94%) and 90% specific (95% CI 68%–98%) for neoplasm, with a PPV of 96% (95% CI 85%–99%) and an NPV of 73% (95% CI 51%–88%).

NIR Imaging Prediction of Postoperative MRI

Of the 16 patients in the SWIG group, 5 had no signs of residual neoplasm, 2 had questionable residual neoplasm, and 9 had signs of residual neoplasm on MR images obtained on postoperative day 1 as well as on follow-up imaging up to 12 months later. Post-resection NIR imaging data from 12 patients were analyzed, since for 4 of the patients, the NIR imaging system was unavailable for the final imaging after resection. Overall, NIR imaging demonstrated residual fluorescence for 8 of the 9 patients with definite residual neoplasm (1 false negative) and 2 of the 3 patients with no residual neoplasm (2 false positives). An example of a false positive is shown in Fig. 5, with a true negative in Fig. 2.

Of the 23 patients in the OTL38 group, 15 had no signs of residual neoplasm on postoperative MRI follow-up. Of the 9 patients with FRα-overexpressing adenomas, 4 had no signs of residual neoplasm on postoperative MRI. In these 4 patients, no NIR fluorescence was detected in the surgical bed after resection, immediately prior to closure (0 false positive). In 4 of the 5 patients with residual neoplasm on postoperative MRI, NIR imaging immediately prior to closure revealed areas of fluorescence in the surgical bed in locations corresponding to the residual neoplasms (1 false negative). In adenomas without FRα overexpression, no significant fluorescence was detected regardless of operative outcome.

Rationale for NIR Fluorescent Molecular Imaging

Intraoperative fluorescence imaging is an attractive option for visualizing neoplastic tissue, as it offers real-time information without significant additional cost or surgical time. NIR fluorophores have 2 main advantages over 5-ALA and other visible-light fluorophores: enhanced tissue penetration by the fluorescence (~10 mm vs μm scale for 5-ALA) and lack of autofluorescence from surrounding benign parenchyma.³⁹ Thus, NIR fluorescence imaging has the potential for improved SBR compared to visible light fluorophores; in addition, NIR fluorescence can be detected through normal dura, which is unlikely to occur with visible-light fluorophores. In this study, we demonstrate high SBR > 3–4, even when using a 4-mm endoscope. The physical constraints of a 4-mm endoscope have not been discussed in this analysis, but suffice it to say that excitation light power and the capture of photon energy on NIR sensors is less optimal with constrained 4-mm endoscopes compared to 10-mm laparoscopes/thoracoscopes or exoscopes.³¹

SWIG Is Highly Sensitive for Neoplasm

Indocyanine-green (ICG) is an FDA-approved NIR dye (peak emission 835 nm) that has been used for decades in vascular surgery for real-time vasculature imaging. When infused at a higher dose 24 hours prior to surgery, ICG accumulates in neoplastic tissue. Our group has demonstrated that SWIG can be used for various intracranial tumors, including metastases, glioblastomas, and meningiomas. It is hypothesized that SWIG works through the enhanced permeability and retention (EPR) effect, which stipulates that enhanced vasculature permeability allows normally impermeable particles to diffuse into and be retained by tumor tissue, whereas normal tissue with less permeable vasculature will show decreased retention of these more macroscopic materials (Fig. 6A).^10,27,28 Thus, the very high dose of ICG administered 24 hours prior to surgery allows microaggregates of ICG to diffuse into and be retained within tumors with leaky vasculature, permitting real-time intraoperative visualization of the fluorescence. Our group previously demonstrated that there is a strong correlation between gadolinium enhancement of tumors on MRI and intraoperative SWIG fluorescence intensity, further supporting the EPR effect’s role in SWIG.²⁴ Since pituitary adenomas are considered hypo-enhancing relative to the normal pituitary gland, our initial hypothesis for this study was that SWIG would not provide sufficient ICG accumulation for visualization.³⁴ We were thus pleasantly surprised to find in this study that ICG did accumulate with high sensitivity. We speculate that pituitary adenoma tissue is sufficiently abnormal in its clearance mechanisms that ICG accrues over 24 hours.

In our previous publications, we showed that SWIG is highly sensitive for intracranial neoplastic tissue, with a sensitivity > 85% for detection of the gross tumor and margin specimens.^22–24 However, SWIG has low specificity (25%–38.9% in our prior studies), due to a high incidence of false-positive signals. Similarly, in this study, SWIG had 100% sensitivity, 29% specificity, 82% PPV, and 100% NPV for pituitary adenomas. The surgeon’s impression, for comparison, was 88% sensitive and 90% specific, with 96% PPV and 73% NPV. Similarly, in the final NIR imaging prior to surgical closure, SWIG demonstrated a falsepositive signal in 2 of 3 patients with no residual tumor on postoperative MRI. The low specificity and PPV of SWIG is thus interpreted cautiously, with the recognition that enhanced vasculature permeability is not a specific finding in tumors; inflammation and tissue injury can also lead to increased permeability, potentially causing false-positive signals.¹⁴ Indeed, in unpublished data, the senior surgeon has also recognized that while normal pituitary gland may not demonstrate significant ICG accumulation, the pituitary stalk and pars nervosa does often demonstrate nonspecific ICG accumulation. Biopsy of these areas has been limited by the unwillingness of the senior surgeon to harvest potentially normal and highly functional neural tissue.

SWIG’s low specificity and PPV have implications regarding its utility. In many surgical oncology subspecialties, surrounding benign tissue is often sacrificed during resection of tumors, as the benefits of maximal resection usually outweigh the disadvantages of the iatrogenic damage. In neurosurgery, however, supramarginal resection is not considered standard-of-care, and thus, a highly sensitive and specific method of tumor identification would benefit neurosurgeons. We do not recommend chasing after all areas of NIR fluorescence due to SWIG’s lack of specificity. We believe, however, that SWIG does have the potential to complement the surgeon’s armamentarium due to its high NPV. In our study, the NPV for the surgeon with white light alone was 73%, indicating that in some cases benign tissue was considered to be neoplastic and resected. SWIG, on the other hand, demonstrated 100% NPV; thus, areas that do not fluoresce with SWIG are likely to be benign tissue and should not be resected. By combining the surgeon’s high PPV and SWIG’s high NPV for distinguishing neoplastic from benign tissue, neurosurgeons could avoid resection of normal tissue while still aiming for gross-total resection. Furthermore, we have demonstrated here that SWIG is widely applicable to all pituitary adenomas, as the preoperative hormonal status of the patient did not affect the NIR signal or the sensitivity, specificity, PPV, or NPV of intraoperative imaging with SWIG.

Hence, we recommend performing surgery based on white light visualization, and using SWIG for confirmation of degree of resection. If there is no residual NIR fluorescence, the surgeon can be very confident that resection has been thorough. In contrast, however, if there is residual NIR fluorescence, the surgeon must interpret this finding with extreme caution, recognizing the lack of specificity of SWIG.

Folate-Receptor–Targeted Imaging Is Highly Sensitive and Specific for Neoplasm

OTL38 is a folate analog linked to an ICG analog that binds and is internalized by cells that overexpress folate receptors (Fig. 6B).^19,42 Folate is a crucial vitamin for cellular growth, replication, and turnover. Folate receptors exist in 4 forms (FRα, FRβ, FRγ, and FRδ), and FRα has been reported to be overexpressed in ovarian, lung, and breast carcinomas.^1,19,36,40 In a series of pioneering studies, it was further shown that nonfunctioning pituitary adenomas overexpress FRα.^7,8 Folate receptors have been successfully targeted for molecular imaging in single-photon emission computed tomography (SPECT), suggesting that they are valid cellular targets.¹¹ Thus, we decided to attempt folate-receptor–targeted NIR imaging in pituitary adenomas, using OTL38.

In our group of 14 nonfunctioning adenomas, OTL38 was less sensitive than SWIG (75%) for detecting neoplasm, but more specific (100%), with a PPV of 100% and an NPV of 62%. The major reason for the reduced sensitivity and NPV was that 5 of the 14 nonfunctioning adenomas in our study did not overexpress FRα on postoperative immunohistochemistry. Since OTL38 can only bind cells that overexpress FRα, NIR signal was predictably unreliable in those 5 cases. When analysis was limited to the 9 FRα-overexpressing adenomas, the sensitivity, specificity, PPV, and NPV were all 100% with OTL38.

Overall, OTL38 imaging demonstrated a superior ability for tissue identification compared to both the surgeon’s impression alone and SWIG, with higher specificity and PPV. This makes intuitive sense, as OTL38 was designed to specifically bind FRα rather than rely on the nonspecific EPR effect, like SWIG. While the NIR signal-to-background ratio (SBR) is lower with OTL38 than with SWIG (3.2 ± 0.52 vs 4.1 ± 0.72, p = 0.003), this difference may not be clinically significant, as our experience shows that any tumor with SBR > 2.5 can easily be visualized. Table 4 summarizes the important comparisons between SWIG and OTL38.

However, precisely because OTL38 specifically binds FRα, its application is limited to certain tumor types. We have previously demonstrated the utility of SWIG in metastases, meningiomas, and gliomas, and now in pituitary adenomas, both functioning and nonfunctioning.^22–24 Conversely, OTL38 imaging, as explained above, was effective in only 9 of the 23 patients who received OTL38. As suggested by Evans et al. and confirmed in our previous study, only nonfunctioning pituitary adenomas overexpress FRα (64% in our series).^7,8,21 While FRα-targeted tomographic imaging with SPECT is capable of estimating FRα expression levels preoperatively, the cost-to-benefit analysis of preoperatively scanning all patients is unclear.¹¹ Instead, we have data³ suggesting that initial NIR SBR of > 2.5 upon tumor exposure may predict FRα overexpression with near-100% probability; thus, a reasonable and affordable approach to using OTL38 clinically might be to administer it to all patients with nonfunctioning adenomas and determine intraoperatively whether each adenoma overexpresses FRα to decide if OTL38 fluorescence would be reliable for that surgery. If the initial SBR prior to resection is < 2.5, resection should proceed using conventional techniques without NIR imaging.

Limitations

There are currently 2 major limitations to intraoperative NIR imaging. First, the fluorescent tissue must be in the line of sight of the endoscope in order to be excited and visualized. With the VisionSense Iridium system currently lacking angled endoscopes, we have discovered that it is difficult to visualize residual tumor around the cavernous sinuses. We believe this explains the one case in which OTL38 fluorescence failed to predict the resection outcome in a FRα-overexpressing adenoma, as that patient’s residual tumor abutted the cavernous sinus. We expect that the addition of an angled endoscope will address this limitation. In craniotomies using exoscopes, this is not an issue.

A second limitation is the low specificity of NIR imaging using SWIG. While this is expected due to its mechanism, the reduced specificity limits SWIG’s reliability. Potential solutions are discussed below.

Future Direction

Advancement in NIR imaging requires the development of both novel fluorophores and advanced imaging modalities. While nonspecific dyes like ICG are widely applicable and have the potential to complement the surgeon’s eyes and hands during surgery, appropriate molecular targets for targeted imaging should also be developed, as targeted imaging is much more specific and could help reduce iatrogenic neurological damage.

Furthermore, our group and collaborators are currently working to develop a tomographic NIR imaging system using spatial frequency domain tomography. Such tomographic imaging, like MRI, would provide depth information and allow the surgeon to assess the thickness of any fluorescent areas. We believe this would be useful in reducing false-positive signals from NIR fluorescence by allowing surgeons to distinguish true residual neoplasm from superficial spillage of fluorophores.