What you see and what you don't - Utility and pitfalls during fluorescence guided resections of gliomas using 5-aminolevulinic acid
Correspondence Address: Source of Support: None, Conflict of Interest: None DOI: 10.4103/0028-3886.236998
Source of Support: None, Conflict of Interest: None
Keywords: Combined modality, fluorescence, intraoperative imaging, utility
The use of aminolevulinic acid (ALA)-based fluorescence to guide resection (ALA-FGR) has become established as a standard of care in surgery for malignant gliomas. Despite its cost implications and the recent re-emergence of other forms of FGR, ALA-FGR remains the preferred technique because of its high specificity for tumour cells., Nonetheless, the technique remains subjective and there remain concerns regarding interpretation of the visualized fluorescence beyond the unequivocal strongly fluorescing tumour core. Practically, many more factors play a role in deciding the completeness of resection (and particularly the resection of all fluorescing tissue). We report our experience with this technique focussing on the challenges, limitations and overall results in our practice.
This was a retrospective analysis of a prospectively maintained database approved by the Institutional Review Board (IRB). Fifty consecutive patients undergoing fluorescence-guided resections (FGR) for lesions that were suspected on radiology to be malignant gliomas were included in this study. Patients were offered FGR as part of routine standard of care when they presented with magnetic resonance (MR) scans that were suggestive of a high grade glioma, if the lesions appeared potentially resectable, patients could afford it and/or if the patients specifically demanded it. Additionally, some cases were operated under fluorescence guidance as part of another ongoing IRB-approved study. All patients were pre-medicated with steroids (dexamethasone 4mg three times a day for at least 2 days) and then administered ALA (Gliolan, Medac GmbH, Germany, @ 20mg/kg body weight dissolved in 50ml potable water) orally, approximately four hours prior to induction of anesthesia. Intraoperatively, a suitably equipped operating microscope (BLUE400, OPMI Pentero, Carl-Zeiss GmbH, Oberkochen, Germany) was employed to visualize the fluorescence. Standard precautions were taken during and after the surgery to minimize the chances of photosensitivity. Surgery proceeded under both conventional white light microscope illumination and the BLUE400 module, switching back and forth as required.
Clinico-radiological and intraoperative details were recorded prospectively in a proforma which was then reviewed for this analysis. Tumours were classified as potentially resectable or unresectable. A few tumours, not amenable to gross total resection, were also offered ALA-FGR either because the patients insisted on the resection technique being a part of the operative procedure or as a part of another IRB approved study. Based on the contrast enhancement characteristics on preoperative MR images, lesions were divided into 3 groups - predominantly enhancing, mixed (significant areas without contrast enhancement) and non-enhancing. Based on functional imaging, tumours were further classified as eloquent (close to or involving primary motor, sensory, speech, visual, auditory areas or their deep white matter tracts, corpus callosum, deep nuclei) or as non-eloquent. Intraoperative functional mapping and monitoring were used whenever necessary. ALA fluorescence was intraoperatively characterized based on the intensity [strong (red), faint (pink) and none (blue)] as well as distribution (uniform or patchy). The white light (WL) appearance was also categorized subjectively as normal and abnormal [Figure 1].
Assessment of diagnostic accuracy of the visualized fluorescence
This was performed in a small subset of twelve patients. Biopsies were collected during the resection from various parts of the resection cavity corresponding to different appearances on white light and fluorescence patterns. These samples were individually processed by a neuropathologist (ES). Besides the routine histomorphological analysis, the neuropathologist specifically interpreted the tumor pattern as coalescent tumor (sheets of tumor cells with no intervening normal tissue), infiltrating tumor (dispersed tumor cells admixed with normal tissue), and normal [Figure 2] and [Figure 3]. For each specimen, the white light appearances, fluorescence pattern, and histological category were tabulated. Two-by-two tables were prepared to assess the accuracy of predicting tumor presence using both white light appearances as well as fluorescence patterns independently. Since a vague fluorescence may have a very subjective interpretation leading to a high interobserver variability, we separately assessed the accuracy of strong fluorescence (SF) alone and then clubbed all the fluorescing tissue together (AF). Similarly, the pathological observations were categorized as coalescent tumor versus non-coalescent tumor (this included infiltrating tumor and normal tissue) initially, and then as all tumor (coalescent plus infiltrating) versus no tumor. Sensitivity (Sn), specificity (Sp), positive (PPV) and negative predictive values (NPV) were calculated for both white light appearances as well as the fluorescence pattern.
The extent of resection (EOR) was ascertained based on the postoperative MR whenever available. Absence of all contrast enhancement as well as all unequivocal T2/fluid attenuation inversion recovery (FLAIR) abnormality was considered as complete resection. We did not routinely perform volumetric quantification of the residual tumor. Routine hematological and biochemical investigation reports were reviewed to record any changes perioperatively. Clinical charts were also reviewed to document any neurological worsening, which was classified as transient (if they recovered by the time of discharge) or prolonged (if they persisted till then).
50 consecutive cases of suspected malignant gliomas treated with ALA-FGR were analyzed. The clinical profile of the patients is shown in [Table 1]. Most of the tumours (n = 46, 92%) were considered resectable. These consisted of either predominantly enhancing (n = 28) or mixed enhancing cases (n = 18). Of the remaining four (two mixed enhancing and two non-enhancing tumours), three were rather diffuse and planned for debulking (with ALA-FGR as part of an ongoing study) and one other patient had a diffuse variegated tumour where ALA was used to identify the most malignant region for biopsy during open surgery. A navigated 3-dimensional intraoperative ultrasound system was used in 60% of the cases along with the fluorescence for guiding the resection. Histologically, most of the tumours (n = 44) were glioblastomas. The remaining six consisted of anaplastic oligodendroglioma (n = 3), anaplastic oligoastrocytoma (n = 2) and anaplastic ependymoma (n = 1).
Pattern of fluorescence
Uniform fluorescence was seen in 62% patients whereas it was patchy in 36% patients [Table 1]. There was one case of a mixed enhancing glioma (anaplastic oligoastrocytoma), which did not show any fluorescence intraoperatively. Though a large number of contrast enhancing tumours did show uniform fluorescence, there were still some tumours where there was patchy fluorescence [Table 2]. Conversely, almost half of the mixed enhancing tumours showed an intraoperative uniform fluorescence. Further, there were two non-enhancing tumours which showed an intraoperative fluorescence. Amongst the tumors which fluoresced, most showed a strong fluorescence (88%) regardless of whether the distribution was uniform or patchy across the tumour. The chi square test showed that predominantly enhancing tumours were more likely to show uniform fluorescence at surgery than tumours that had mixed patterns of enhancement on MRI (χ2 – 4.565, P = 0.03).
Overall, 29 biopsies were obtained from twelve patients for this assessment. The accuracy of white light findings and fluorescence with respect to the histology is shown in [Table 3]. For detecting the coalescent tumor, the white light (WL) was more sensitive and had a better NPV, but SF was more specific and had a higher PPV. SF had maximum Sp and PPV (both 100%) for detecting any tumor (higher than for coalescent tumor) due to the fact that there were areas of coalescent tumor which actually had only vague fluorescence. On the other hand, the Sn and NPV of WL dropped for detecting any tumor (versus the coalescent tumor) due to the fact that there was a WL normal appearing area which contained infiltrating tumor.
Residual fluorescence and its causes
Residual fluorescence was documented in 27 of the 50 cases overall (54%), and even in the resectable group of tumours, 26 of the 46 (55%) cases showed a residual fluorescence [Table 4]. The commonest reasons for leaving behind fluorescing tissue were the involvement of eloquent areas (13 cases) or important vessels (3 cases) [Table 5]. In the remaining 10 patients, diffuse ependymal fluorescence which could not be resected was the cause for residual fluorescence in 9 patients, and excessive tumour vascularity and bleeding resulted in abandoning further debulking in one more case. Interestingly, in tumors close to eloquent areas, the residual fluorescence was not always near the eloquent area [Table 5].
Extent of resection and correlation with intraoperatively adjudged resection status
Postoperative MR was obtained in 46 of the 50 cases. In 4 patients, only a CT scan was performed. All four patients had a resectable tumor, but MR could not be obtained in them due to logistic issues. The extent of resection was calculated in only these 46 cases. Complete resection was obtained in 30 cases (with a gross total resection [GTR] rate of 65%). This included 28 potentially resectable tumors leading to a GTR of 67% in this subset of patients (n = 42). The remaining two cases of GTR were interestingly seen in the subset of “unresectable” tumors.
The correlation of intraoperatively documented residual disease (based on the surgeon's evaluation with fluorescence and ultrasound) and postoperative MR defined residual disease is shown in [Table 6]. Intraoperatively, 15 patients were deemed to have a complete resection (no fluorescence and no residue on ultrasound [US] imaging). MR imaging confirmed the absence of residual disease in all but one of these cases. In addition, there was no residual fluorescence in 8 more cases. However, in these patients, non-fluorescing residual disease was judged to have been left based on the findings of intraoperative ultrasound. MR imaging eventually showed a residual tumor in 5 patients, but could not pick up any residual disease in the other 3 patients. Thus, relying purely on intraoperative fluorescence would have missed a total of 6 cases with residual disease (though using another complementary technique like ultrasound would have reduced this to just 1 patient). On the other hand, residual fluorescence was documented in 23 cases. Of these, MR imaging showed a residual tumor postoperatively in only 10 cases, whereas it was thought to be a complete resection in the other 13 patients. Of these 13, in 5 cases, the fluorescing residue was along the ependyma, whereas in the remaining 8, it was in the vicinity of eloquent areas or important blood vessels. These 8 patients were thus missed by the postoperative MR imaging.
Two patients developed transient abdominal cramps after consuming the ALA (prior to surgery). This pain settled spontaneously and further surgery and postoperative period was uneventful. Leucocytosis (a white blood count >16,000/cc) was documented in 30 patients within the first 24 hours. Two patients had thrombocytopenia (in one, on the postoperative day 1, and in the other one, on day 7). Four patients had a mild (asymptomatic) derangement of liver function tests on postoperative day 14. None of our patients suffered from photosensitivity due to the dye.
Seven patients (14%) had transient worsening of their neurological status, which recovered by the time of discharge. Eleven (22%) others had prolonged neurological worsening after surgery. All of these had tumors in the eloquent regions. There was no death in the immediate postoperative period.
FGR has greatly improved our ability to visualize and resect malignant gliomas. With evidence supporting the role of extended resections in improving survival in general, there is no doubt that using adjuncts like FGR is likely to benefit patients even though there may be no direct evidence to support its survival benefit., Our study showed that in the day-to-day practical scenario, ALA FGR is very useful. Using ALA-FGR, we were able to achieve a reasonably high GTR (66%) in malignant gliomas with no added morbidity. However, a few caveats need to be borne in mind in order to optimally utilize the benefit of this technique.
Tumor enhancement and ALA-FGR
Typically, FGR using ALA has been recommended in contrast enhancing gliomas assuming that a breach in the blood-brain-barrier is a pre-requisite for the technique to work. Contrast enhancement has been shown to correlate with tumor fluorescence., It is also known that fluorescence is stronger (red) in high-grade areas of the tumor, correlating with a higher tumor cell density and proliferative indices.,,, Considering the fact that a sizeable proportion of high-grade gliomas may not show enhancement and still fluoresce, there may be other biological mechanisms at play. Indeed, FGR may have a role to play in non-enhancing malignant gliomas too. An indirect corroboration of this phenomenon is the fact that ALA fluorescence has been shown to extend beyond the contrast enhancing tumor zone. In our present study too, strong and diffuse fluorescence was seen even in many of the patchily enhancing and even some non-enhancing tumors. Nonetheless, with the existing technology, contrast enhancement still remains a reasonable surrogate marker to predict the intraoperative usefulness of ALA-FGR.
Accuracy of FGR
The fundamental basis of ALA-FGR is its proven accuracy in predicting the presence of a malignant tumor. Indeed Stummer et al., first reported a very high specificity (100%) of the procedure for detecting a tumor. The same study also reported the sensitivity as being 85%. The slighty lower sensitivity value was attributed to areas of necrosis and infiltrating tumor which did not fluoresce. This report did not differentiate between various patterns of fluorescence or the pattern of tumor at histology. Diez-Valle reported a series of 36 cases where pathological correlation was done. They too reported that a strong fluorescence has a 100% PPV for detecting a glioblastoma. Our present study similarly shows a very high Sp and PPV whenever a strong fluorescence (SF) is present, for detecting any tumor (100%). However, the SF in our study was also seen in samples which had an infiltrating tumor, resulting in a lower Sp and PPV for SF when only coalescent tumor was considered (92%). Interestingly, another study by Stummer et al., showed that SF was more often associated with coalescent tumor and increased cellularity. However, this correlation was not an all-or- none phenomenon. Only 54% of SF tissues showed a coalescent tumor and fewer (44%) had high cellularity. On the other hand, Utsuki et al., have reported higher false positive rates (and hence lower specificity) for fluorescence. Colditz and Jeffree reviewed the studies which have attempted a pathological correlation with the pattern of fluorescence. They reported Sn, Sp, PPV, and NPV of “any fluorescence” and “strong fluorescence” as being 91%, 59%, 85%, 71% and 77%, 92%, 92%, 79%, respectively, in predicting the presence of a “high-grade glioma”. Our present findings with respect to the prediction of a coalescent tumor (which would correlate with a high-grade tumor) are very similar to these figures and hence provide validity to the results. In another study in malignant gliomas, Hefti et al., reported the sensitivity and specificity for strong fluorescence, vague fluorescence, and white light as being 100 and 98%, 85 and 76%, and 68 and 66% respectively. Here again, the pattern of tumor was not separately defined (coalescent or infiltrating). However, these authors mentioned that vague fluorescence is more often seen in the transition zone where the infiltrating tumor may affect the accuracy of interpretation. The authors proposed a computerized algorithm and a system of optical quantification in order to objectively interpret the fluorescence. Sanai et al., have shown that the sensitivity of intraoperative detection of fluorescence can be improved using confocal microscopy. Though strong visualized fluorescence has a high predictive value, it still is not without limitations. It must be realized that strong and vague fluorescence are very subjective interpretations. The operating surgeon must interpret these patterns of fluorescence cautiously. Vague fluorescence is especially difficult to interpret, as seen in our study too, and any decision based on this finding alone may be hazardous. Therefore, till such time that more accurate means of quantifying visualized fluorescence are available at the disposal of the surgeon, one must tread with measured aggression. Fluorescence is preferably to be used in combination with other intraoperative adjuncts if available, as well as with functional monitoring in and around eloquent areas. Conventional radiological intraoperative imaging (using US and MR), which provides a cross-sectional imaging, complements the information available from fluorescence imaging (which remains a surface phenomenon), and could be used beneficially in combination. Eyupoglu et al., have recently shown that an intraoperative MRI can improve the GTR rates achieved by ALA-guided surgery alone. Tsugu et al., have also shown that using intraoperative MRI along with fluorescence guidance is better for non-enhancing tumors where the fluorescence may be vague or absent. In our experience, intraoperative ultrasound has been a very useful complementary adjunct. As respecting and preserving functional areas is of paramount importance, functional monitoring combined with fluorescence guidance is advisable in the vicinity of eloquent areas. Neurosurgeons, especially those treating brain tumors, should be conversant with and have at their disposal all available adjuncts to customize and optimize surgical resection.
Though the aim of surgery during ALA-FGR is to remove all of the fluorescing tissue, this is not always possible. It is important to acknowledge this fact and employ neurosurgical “common sense” rather than chase all the fluorescing tissue. In our experience, there was residual fluorescence at the end of surgery in more than half (54%) the cases. Similarly, in a recent report by Stummer et al., almost 70% cases had residual fluorescence. Even in the group of tumors which were deemed to be resectable, fluorescing tissue was left behind in over half (55%) of the cases. In such cases, the major cause for residual fluorescence is its proximity to crucial neuro-vascular structures (16 of 26 cases). The guiding principle of “safe maximal resection” takes precedence and fluorescing tumor needs to be left behind in the interest of safety. This is particularly true for what is loosely labelled as 'faint or vague fluorescence', which may be a very subjective interpretation. Besides, there is a sizeable group of patients where diffuse fluorescence is encountered along the ependymal lining (9 of 26 cases). In our earlier experience, this does not correlate with pathological tumor and its clinical significance is equivocal. This residual fluorescence need not always signify the presence of a residual tumor and needs to be further studied carefully in the future. Further, as we have shown and has already been reported earlier, the presence of fluorescence predicts the presence of residual tumor well (high PPV); however, the absence of fluorescence does not rule out residual tumor (indicative of a poor NPV). This is especially true for the non-enhancing component of the tumor. Thus in five cases, there was no residual fluorescence, but US picked up the tumor, which was then resected. No single modality can be considered sufficient by itself and it is best to combine adjuncts in order to obtain the maximum benefit. Recent evidence suggests that resecting the tumor beyond the contrast enhancing component may improve oncological outcomes and reinforces the need to be better equipped intraoperatively to achieve this goal. It is important to note that the high PPV of fluorescence in detecting pathological residual tumor is not the same as the correlation of residual fluorescing tumor with postoperative contrast enhancing tumor. In 35-45% cases where residual fluorescence was left, MR still showed complete resection of the enhancing tumor (CRET)., In our study, even after accounting for the T2 tumor component, 52% of cases with residual fluorescence had a complete resection, as adjudged by the MR imaging. This reasserts the fact that ALA induced fluorescence depicts the presence of the tumor beyond the contrast-enhancing zone (including the T2 component), as has been shown earlier. Under favourable circumstances (the presence of a non-eloquent location and in non-ependymal areas), resection of the tumor in these areas could also be facilitated. However, as already stated, a combined modality approach (using ultrasound, for example) may help to maximize the benefit. Though we did not perform a volumetric assessment of the residual tumor, we did include T2/FLAIR signals in the assessment of residual disease and this could explain the slightly lower GTR rates (66%) that we obtained compared to the more recent publications, which have been based the extent of resection of the contrast-enhancing component of the tumor., Moreover, our experience has also confirmed the minimal morbidity associated with this technique, as has also been shown in other larger studies.,,
The present study is a small one (50 cases). Nonetheless, it is one of the first such reported experience from the Indian subcontinent., It reinforces the earlier reported benefits and the accuracy of this novel technique and also highlights some important practical limitations and precautions that must be adhered to during the performance of FGR. The cost of the agent (ALA) and the microscope modification of the tumor remain major challenges limiting the wider availability of this technique in this geographical region.
ALA-FGR is a very useful intraoperative adjunct in malignant glioma surgery. Fluorescence does not always correlate with contrast enhancement and may be seen in non-enhancing tumor components, too. For this same reason, postoperative MRI (especially contrast enhancement only) may underestimate the true extent of residual disease. Objective evaluation of the T2 residue on MRI would be a better and more reliable marker of residual disease status. Fluorescence shows the intraoperative tumor extent beyond the white light and even in the MR delineated areas and certainly helps the surgeon to be better guided. However, the fluorescence may extend into vital eloquent areas (where resection is hazardous) or into non-specific ependymal areas (where resection may not be of any added benefit) and hence the surgeon has to exercise judicious decision-making rather than blindly follow the fluorescing tissue. If the inherent limitations of the technique are paid heed to, it is safe, and radical resections can be obtained without increased neurological morbidity.
Financial support and sponsorship
This study was partly funded by an intramural grant from Tata Memorial Centre, Department of Atomic Energy- Clinical Trial Centre and Directorate of Biotechnology, India.
Conflicts of interest
There are no conflicts of interest.
[Figure 1], [Figure 2], [Figure 3]
[Table 1], [Table 2], [Table 3], [Table 4], [Table 5], [Table 6]